Quality theory and methodology

The objective of pure price indexes is to measure price change over time to constant quality. This is achieved by re-pricing an identical basket of products each period and ensuring the basket of products is unaffected by quality and quantity changes. This is often referred to as pricing to constant quality and an important element of the day-to-day role of price statisticians.

The concept of quality is based on the notion of utility to the purchaser. Quality change is measured by reference to the expected value of the changes to the purchaser. While it is not always possible to achieve this in practice, it is the principal guideline in making decisions concerning quality change.

In economic theory it is generally assumed that whenever a difference in price is found between products which appear to be physically identical, there must be some other factor, such as location, timing, conditions of sale etc., which is introducing a difference in quality. Otherwise it can be argued that the difference does not exist, as rational purchasers would always buy the lower priced products and no sales would take place at the higher price.

However, underlying the economic theory are some strong assumptions which rarely hold true in the marketplace. The key assumption behind “different price means different quality” is that purchasers have
perfect information and that they are free to choose between products offered at different prices.

In most markets, purchasers do not have perfect information about existing price differences and may therefore inadvertently buy at higher prices. While it is true that most purchasers will search out lower prices, costs are incurred in the process. The lack of information about price differences may result in search costs being greater than price differences, in which case rational purchasers may be prepared to accept the risk that they are not in fact buying at the lowest price. The existence of this imperfect information in the marketplace is evident from the number of situations where buyers or sellers negotiate over prices.

Even when purchasers are well informed, in practice they are not always free to choose the price at which they purchase. This situation arises because of price discrimination, whereby the seller is in a position to charge different prices to different categories of purchasers (for products that are otherwise identical). Price discrimination is a common practice in the marketplace as it enables sellers to increase revenues and profits.

Complicating this observation is the difficulty that arises when purchasers can resell amongst themselves (that is, purchasers that buy at the lowest price can resell products to other purchasers). Under such circumstances price discrimination is less likely to occur. While this circumstance can occur for the sale of most products, this situation rarely arises for services, since it is usually impossible to resell services. Price discrimination is frequently practised in many of the transportation and business service industries for this very reason. Therefore, when different prices are charged to different purchasers it is essential to establish whether there are in fact any quality differences associated with the lower prices.

The applicability of the underlying economic theory is tested when differences in price arise because there is insufficient supply. Such a situation typically occurs when there are two parallel markets. For example, there may be a primary domestic market and a secondary market for imports. If the quantities available in the domestic market are limited, there may be excess demand so that supplies may be imported. As a result, the price on the secondary market will tend to be higher. It is also possible that products from the secondary market are not only more expensive but of different quality.

Therefore, prices statisticians are faced with a contrast between theory and practice: theory indicates that a difference in price means a difference in quality, but a difference in price may also arise due to lack of information, price discrimination by customer type, supply constraints and/or the existence of parallel markets. Thus, the existence of different prices does not always reflect corresponding differences in the qualities of the products.

The term ‘quality’ embraces all those characteristics in a product that the purchaser values or from which it derives utility. Therefore, the problem is to identify those characteristics that purchasers value, to make an estimate of the value of those characteristics and to measure the change in those characteristics embodied in the product so that their effect can be removed when calculating price movements. When used in this context, ‘quality’ encompasses all attributes of a product, including quantity.

Sets of products are available in the marketplace with physical characteristics which differ from each other. For example, potatoes may be old or new, red or white, washed or unwashed, loose or prepacked. Loose unwashed Russet Burbank potatoes (used for French fries) are a different quality of potato to washed, prepacked Atlantic potatoes (used for crisps).

When sets of products are sufficiently similar to be considered the same generic type of product (such as a potato), but have sufficiently different characteristics that make them distinguishable from each other from an economic viewpoint, the products are said to possess different qualities.

Not all differences in quality are attributable to differences in physical characteristics of products. Products with identical physical characteristics delivered to different locations, or at different times, are considered to have quality differences. Purchasers situated in one location frequently have different marginal utility from that of purchasers in other locations; hence, different locations may result in different qualities.

Identical products provided at different times of the day (or year) must be treated as different qualities. An example of such a quality difference occurs with the supply of electricity. Electricity provided at peak times is considered to be of a higher quality than that provided at off peak; the very fact of a peak time shows that purchasers of electricity attach greater utility to the provision of electricity at these times.

Quality differences may also be determined by a range of other non-physical attributes. Quality may be determined by conditions of sale, presence of free after sales service, guarantees for durable products, inclusion of delivery, methods for payment, and so forth.

Frequently the precise product priced in one period is no longer available in the next period because either there has been some change in the characteristics of the product or else something new has taken its place. For price index purposes it is necessary to devise techniques to identify quality differences and eliminate their effect on prices from the calculations of price change for inclusion in the index.

The objective of pure price indexes is to measure price change over time to constant quality. This is achieved by re-pricing an identical product or service each period and ensuring the product or service is unaffected by quality and quantity changes. The dynamic nature of products and services, where differentiating and improving quality of products and services is a key element, pricing to constant quality cannot be achieved without the application of quality adjustments. A failure to price to constant quality would result in a pure price index that reflected price change and quality change.

Adjusting for quality prevents distorted price statistics that can occur from unsuitable quality adjustments or incorrect application. Biases can arise from the inability to account for changes in quality over time.

Quality adjustments create conceptual and practical challenges, the importance of correct procedures ensure that price statistics measure pure price change.

The importance of pricing to constant quality is evident when the primary purpose of the Producer and International Trade Price Indexes is to support the calculation of volume measures in the Australian National Accounts and Balance of Payments. Pricing to constant quality results in the quality change being correctly reflected in the volume measures when deflation of current price estimates occurs.

The real world of economic transactions is ever changing and dynamic. Frequently the product priced in one period is no longer available in the next period because either there has been some change in the characteristics or something new has taken its place. Specific varieties of products regularly appear then disappear. New products can appear because of advances in technology, making the production of these new products possible.

Failure to account for quality changes would introduce a bias into the price index. Thus, if the qualities of products being compared are not identical, there are effectively two options:

• to adjust the observed price of the old quality for the change in quality which has taken place (referred to as a quality adjustment) or
• to treat the two qualities as if they were two separate products and obtain their prices in the periods in which they were not collected.

The problem facing price statisticians is isolating and quantifying the direct effects of changes in the quality on products they are pricing in the fixed basket, to achieve an index of pure price change. However, the ABS has several options for quality adjustment available for index managers and further information on the methods are available below.

Quality adjustment is defined as making a change in the price or price movement that accounts for the change in quality that affects the utility of the product.

In compiling the Producer and International Trade Price Indexes, five situations and treatment methods are clearly defined when a product changes:

• Overlapping sales - where there is at least one period when both qualities are on sale in the market at the same time
• Non-overlapping sales - where one quality is replaced by another of different quality, but the two qualities have not been available in the market at the same time
• Component approach - where there are some changes in the composition of a particular quality
• Hedonics - where the different qualities are assumed to be functions of certain measurable characteristics or
• Not directly comparable - where qualities are different, but no information exists to allow an explicit quality adjustment to be made.

Overlapping sales arise where a particular product being priced is no longer available in the market place from one period to the next, but there is another similar product that has been, and continues to be, available in the same market as the initial product and is expected to be a substitute once it is discontinued.

In this situation, provided the two products were sold side by side for some time in the same market and both were sold in reasonable quantities, the approach is to collect prices for both products at the one date and to assume that the difference in prices represents the difference in quality between the two. The assumption is that the market has adequate knowledge of the qualities and prices of each product and that the difference in price is regarded by them as a reasonable measure of the difference in quality. The second product is then substituted for the first using the technique of splicing price series, as illustrated below:

Example: consider two harvesters for sale. Harvester A is for sale in period 0 and period 1. Harvester B is for sale in period 1 and period 2. The two harvesters are both for sale during period 1, in which case there is an overlap of the two products. The products are considered to be of different qualities.

The price movement reflected in the index from period 0 to period 1 is the movement in the price of Harvester A (6.3%). The price movement from period 1 to period 2 is based on Harvester B (3.2%), which will be priced in subsequent periods to replace Harvester A. The difference in price between Harvester A and Harvester B has been eliminated through the process of splicing the new price series to the old price series.

An equally applicable interpretation of this process is to consider the comparison of prices between period 1 and period 2. In period 1, Harvester A is priced at $85,000. In period 2, Harvester B is priced at $98,000, and this is interpreted as:

• a quality change of $10,000, from $85,000 to $95,000
• a price change of $95,000 to $98,000.

In some cases, even with overlapping sales, simple splicing of the price of the new specification to the existing price series is not a satisfactory way of eliminating changes in quality. This situation occurs, for example, when the price of a new model reflects not only the extent of modifications but also a degree of price change, upwards or downwards, for reasons quite distinct from these modifications. In these circumstances, a simple splicing of the old and new prices would eliminate the elements of pure price change as well as the elements of change in quality. In such cases, it is necessary to assess the degree of pure price change involved and to ensure that this is reflected in the price series after splicing.

Where two qualities are not sold in the marketplace at the same time, it is necessary to implement indirect methods of quantifying the change in quality. This circumstance arises frequently for durable products, such as motor vehicles, white goods, home entertainment products and so forth, where producers cease all production of the superseded model when a new model is introduced. In these cases, it is necessary to estimate the relative prices of the old and new models, had they been sold in the market at the same time. The estimated relative prices then give an indication of the measure of the relative qualities.

In many circumstances the difference between the old and new products is a matter of size or dimensions. For example, an 80g jar of instant coffee is replaced with a 100g jar. In such cases the difference in price is readily determined by considering the per unit price; in the case of instant coffee, the change in price per gram would yield the quality adjusted price change¹. 

This process is equivalent to considering the difference in size as a difference in quality, and then explicitly pricing this difference. In the case of the coffee jar, the process would be to determine the price of the extra 20g. This value would then indicate the quality change between the smaller and the larger jar. In practice, this value of the quality difference is used to quality adjust the previous period price, such that the products in both the current period and the previous period are of the same quality.

These data may be interpreted as:

• a quality change of $1.05, from $4.20 to $5.25, representing the addition of an extra 20g of coffee; and
• a price change of $-0.25, being the fall from $5.25 to $5.00.

The method used to adjust for changes in the composition of a quality is to identify the quality difference and place a value on that difference. Frequently the composition of a particular product changes because of the use of different materials or the addition or deletion of particular features.

An example would be a change in the wool/synthetic mix of a yarn. In such cases, the technique used to estimate the value to the user of the quality change involves ascertaining the additional cost (or saving) to the manufacturer and examining the prices of broadly comparable products (e.g. yarns containing various proportions of pure wool and synthetic fibres).

Sometimes the modified product differs markedly from the previously priced product. An example of such changes occurs with the change in model for a particular make of motor vehicle. This type of quality change requires the collection of a considerable amount of information and, in some cases, subjective judgement is required to estimate a monetary value by which to adjust the price. The first step is to obtain a full picture of the differences between the old and new models. This is done by:

• obtaining detailed information from the manufacturer or industry associations, such as design and engineering reports
• examining published tests and other comments on the new model in trade publications, magazines, etc.
• physically examining the new model and questioning producers about the nature of the changes.

Having identified the precise differences between the models, the next step is to determine which of these differences represents changes in quality and to estimate the monetary value of each change. Some changes are relatively simple to quantify. Continuing the car example, changing the type of tyres on the new model when both types of tyres are sold separately in the market is readily assessable since the value of the quality change can be assessed as the difference in the selling prices of the tyres.

Other changes require more detailed examination. If we consider motor vehicles again, a new model car may have leather covered seats while the old model has cloth covered seats, in which case the factors that would be considered are:

• the unit cost to the manufacturer of the change
• whether the leather covered seats have been previously available as an option and, if so, what was the price and did a significant number of buyers purchase the option at that time
• the change in comfort and durability of the seats.

When faced with measuring prices for products which undergo rapid quality change, international best practice is to develop hedonic price indexes when suitable source data are available. This is the approach being advocated by international agencies such as the Organisation for Economic Co-operation and Development³, the International Labour Organization and the International Monetary Fund.

A hedonic price index is any price index that utilises, in some manner, a hedonic function. In broad terms, a hedonic function identifies the relationship between the prices of different varieties of a product, such as differing models of personal computers, and the characteristics within them. By comparing prices and features of various computers, a hedonic regression model assigns values to each of the particular features that are identified as price determining (for example, processor speed, memory, disk capacity etc.).

Personal computers are an area of rapid technological change. Products available in the marketplace change frequently as new features are added and existing features improve. For example, the rapid change in hard disk size, random access memory, and clock speed of desktop personal computers is well documented. A further issue is that older models quickly become redundant. The net result of these changes is that over any two periods there are both new products and discontinued products, with the result that comparing like with like becomes difficult. This is of particular concern when it is observed that improved features on later models do not always result in a price rise, or a commensurate price rise that would be observed if the components were bought separately (again, a bigger hard disk drive is an example). The quality adjustment problem is applicable to all price indexes, not just those for personal computers. However, traditional approaches to solving this problem (for example, matched model approaches, explicit quality adjustments, or component level pricing, amongst others) are inadequate for these sorts of products.

The Producer Price Indexes use a form of hedonic index known as the 'consecutive two period chained time dummy double imputation hedonic price index' for use in price indexes for personal computers. This process sees a matched model price index applied for personal computers sold between consecutive periods, combined with a consecutive-period time dummy price index (this is produced by using regression techniques) to measure price changes for both discontinued and newly introduced products.

The double imputation method can best be thought of as a traditional matched model index with an explicit adjustment applied because of both the departure of superseded models and the introduction of new models. A key deficiency of the basic matched model approach is that it makes no provision for systematically including the effects of price and quality changes in models available in the marketplace, and determines price change by only considering those models which appear in the market in both periods of interest. In other words, any improvement in quality associated with the introduction of a new model will not be measured if only matched models are priced.

The double imputation price index counters this deficiency by implicitly imputing price movements for both superseded models and newly introduced models. This is where the term 'double imputation' arises. A hedonic regression model is run on the dataset each period and from this, a price factor is determined for each characteristic of the computer. Whilst this gives the ability to calculate the price for each specification, what is actually used in the imputation, is the time dummy variable. The time dummy variable is representative of the price change between the two periods taking into account the different characteristics of the computer. This is combined with the matched model index to create the double imputation index which will reflect the movement of the whole sample. The index is then considered representative of all transactions, since recently superseded models and new models are included in the determination of price change, in addition to products common to both periods.

Further, the implicit imputation process at the core of this technique uses a hedonic function to adjust for changes in the characteristics of both the new and superseded models; that is, the prices imputed are adjusted for quality change, and hence the resulting index measures pure price change.

The process utilises price data from Australian vendors of personal computers, and so is not only representative of the Australian marketplace but also avoids issues with both exchange rate fluctuations and arbitrarily lagging prices to take account of shipping times etc. The double imputation index uses a hedonic function based on characteristics of personal computers sold in the Australian marketplace, using prices in Australian dollars, and so furthermore does not rely on the restrictive assumptions underlying a universal hedonic function.

Any movements in the double imputation index can be decomposed into the movement due to changes in prices of the matched sample, and the movement due to changes in other products in the marketplace. Movements in the index can be explained in terms of changes in list prices of existing products and changes in quality of new products, and so the resulting measures are easily explainable to users.

Despite the efforts to quantify explicitly differences in quality, circumstances occasionally arise where insufficient information exists to value the differences between two models of a product. In such cases it is still necessary to make a quality adjustment to allow comparison of prices. In the absence of any other information, the strategy employed in the Producer and International Trade Price Indexes is to consider the problem in two parts:

1. If the old model had existed in the current period, how much would the price have changed between the current and previous period?
2. What would the price have been in the current period?

If no other information is available, the quarterly movement in the price of the old model needs to be estimated from the price movements of similar products. In this manner the price movement problem is directly comparable to the imputation and temporarily missing price observations problem described in the Producer and International Trade Price Indexes Calculation in practice section.

Once the price movement has been determined, an estimate of the current period price for the old model can be made by working forwards from the observed previous period price. The new model is then introduced in the subsequent period (with a back price), and any difference at that time between the estimated price for the old model and the (observed) previous period price for the new model is due to the difference in qualities between the two models.

This approach is only used in circumstances where other efforts at quality adjustment have been exhausted, since excessive use would introduce bias into the price indexes. Such biases arise because of the assumptions underlying the application of this technique.

1. Prices for new models move the same as prices for other models - the use of the not directly comparable technique has an implicit assumption that the new model has the same price movement as other models. It is known across some industries that prices are increased when new models are introduced (as a means of increasing revenues and profits), in which case use of this mechanism would result in a downward bias for these types of products.
2. Prices change for reasons other than the introduction of new models - a more severe consequence of the prices move the same assumption occurs for those products where the only price movement occurs when models are changed. In this case, prices for continuing models remain static from period to period. The introduction of a new model would always result in an imputed (estimated) price change of zero - meaning that the index would never change. This may also be interpreted as all differences in observed price being solely attributable to quality change.

The incorporation of new products into an index, results in high quality indexes because it ensures they have a representative sample. New products exhibit different pricing behaviours to established products. Excluding them will result in biasing of the indexes.

The production or importation of a new product causes particular difficulties in compiling indexes and these difficulties arise because:

• new products are difficult to identify when using a fixed basket; in particular, difficulties arise when differentiating new products from improvements to existing products
• measuring price changes for new products has its own range of issues
• incorporating a new product into a fixed basket index requires index restructuring, measurement of value data for the new product and reweighting of the existing index structure.

The key question in identifying new products is differentiating new products from existing products whose quality has changed. A practical definition of a new product is that the new product cannot be effectively linked to an existing product as a continuation of an existing resource base and service flow. For example, the VCR was a completely new product when it was introduced in the 1970s because nothing like it had existed before. On the other hand, the DVD recorder replaced the VCR when it was introduced.

Identifying new products requires regular assessment of the marketplace, involving ongoing liaison with producers, regulatory authorities, and industry associations. The ABS approach this problem using several different methods.

For the International Trade Price Indexes, new products are identified through the analysis of data from the Australian Border Force, and International Merchandise Trade Statistics. These data are particularly useful since they not only highlight the emergence of new products but also the value of sales and purchases, indicating the importance of any new product.

The Producer Price Indexes use a variety of different instruments to detect the emergence of new products. First, questions regarding new products are asked each period in the Survey of Producer Prices. Second, regular contact is made with providers outside the quarterly cycle to assess specifically potential changes in production. Finally, a program of personal visits is made to providers and industry associations.

After a new product is identified, its price for two consecutive periods needs to be determined before it is included in the price index. This requirement ensures that a quarterly price movement can be associated with the new product. Such a requirement means that back prices are sometimes used, where the price in a previous period is supplied in a later period. In all circumstances, it is necessary for the product to have existed in the marketplace sufficiently long enough for a price movement to be determined. Of further concern is that an individual new product will almost certainly represent other such products sold in the marketplace, in which case it is necessary that any such initial price movement is representative of the entire market (for the new product).

A new product, by its very nature, does not belong to the fixed basket of a price index and must be introduced at some point after its arrival in the marketplace. As described above, the bias associated with new products is exacerbated through delays in introducing it into the price index. Yet a new product can only be introduced when both sufficient value data for the product exists and when the price index is reviewed.

Value data (either revenue or expenditure depending upon the nature of the price index) are essential for incorporating any product into a price index. Introducing a new product will also have an impact on the revenue (or expenditure) of other products in the marketplace. Consequently, value data are required for not only the new product but also for other products in the price basket. The introduction of a new product results in a value aggregate (and therefore a weight) being attached to the new product, and a change in the value aggregates of other products in the basket.