CPI Technical Paper

Methodology change for Computers and Printers

Introduction

This document describes changes to the methodology for calculating the Irish Consumer Price Index (CPI) and Harmonised Index for Consumer Prices (HICP). These changes are for Personal Computers and for Accessories for information processing equipment. It describes the methodology change and the motivation for this change.

Details of methodology change

Method up to August 2019

The index for Personal Computers is calculated by pricing the same sample of laptop and desktop computers each month and calculating the relative change in price using a geometric average of the individual price changes.

The index for Accessories for information processing equipment is calculated by pricing the same sample of printers each month and calculating the relative change in price using a geometric average of the individual price changes.

When a computer or printer is no longer available, it is replaced in the sample and the new computer or printer is priced each month after this until it too needs replacing. These replacements happen frequently for computers and printers where there is a rapid turnover of what is being sold.

The problem of how to calculate the change of price between the old and new product when we make a replacement is known as Quality Adjustment. There are different methods of quality adjustment and it is important to choose the method most suitable to the item being priced.

The method used up to August 2019 for computers and printers is called Bridged Overlap. It works as follows:

When a specific computer priced in a retail outlet is no longer available, a replacement must be found to continue the measure of price change. Using the bridged overlap technique, no price comparison is made between the two computers when the replacement is brought into the index. The price change between the two observations is imputed as the average change for the other matched price observations for computers. For a price observation to be brought into the index, it must be available for two consecutive periods. The same method is used for printers.

Reason for changing the methodology

There are two reasons to think that bridged overlap is not the most suitable method of quality adjustment for computers and printers.

1. Computers and printers are often introduced to the market at their full price and leave the market at a sale price. At this point they are replaced in our sample by another computer/printer at full price. This pattern is repeated with the replacement. Using bridged overlap with products exhibiting this pattern can lead to a downward bias in the index.
2. Bridged overlap does not consider the quality changes (e.g. increase in memory size, CPU) that exist between the replaced computer/printer and the replacement.
   A quality adjustment technique is needed that resolves these two issues.

New Method

From September 2019, the quality adjustment technique of Hedonic Regression is used when there is a need to have replacement computers and printers in the sample. The benefit of this method is that it allows comparison to take place between the old and new products and it also allows for an estimate to be made for quality change.

Under this method, a regression model is used to estimate the contributing effect of a variety of characteristics associated with the computer or printer. Examples of such characteristics for computers are CPU and memory. Examples of such characteristics for printers are pages per minute and resolution width. These characteristics serve as the independent variables in the model, while the natural logarithm of the price of the computer or printer acts as the dependent variable. We have three models, one for laptops, one for desktops and one for printers. Using these models, it is possible to quantify the quality change through the difference of these characteristics between the original product and it’s chosen replacement. Once we have quantified the quality change between the old and new products, we can then estimate how much of the difference in price is a pure price increase or decrease.

This method of quality adjustment is designed to address the two issues mentioned in the previous section.

1. If the older computer/printer was at a reduced price when replaced, and the replacement computer/printer is initially at full price, this will now likely be calculated as a price increase once the hedonic quality adjustment has been made.
2. The hedonic quality adjustment takes quality changes into account. For example, if improved computers are being sold at the same price as lower quality computers were being sold previously, this will now result in the index decreasing.

The models relating the prices of computers and printers to their characteristics will only be valid for a limited amount of time, because this is a rapidly changing market. We anticipate that we will update each of our models every year, based on the most recent data for prices and characteristics of computers and printers. In this way we ensure that our quality adjustments for replacements are based on the current market.

Revisions

The change in methodology described here will not lead to any revisions to the CPI or HICP.

Once the CPI indices are published, they are never revised.

The HICP can be revised in the case of a mistake or if there is new or improved information. Neither of these is the case with this change in methodology. See the EC Regulation number 1921/2001 for the HICP revisions policy.

https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32001R1921&from=EN

Further reading

Eurostat HICP Methodological Manual - https://ec.europa.eu/eurostat/documents/3859598/9479325/KS-GQ-17-015-EN-N.pdf/d5e63427-c588-479f-9b19-f4b4d698f2a2

Chapter 6 is on Replacements and quality adjustment. It explains bridged overlap, hedonic regression and many other quality adjustment techniques.

CSO Paper presented at UNECE – Quality Adjustment in the Irish CPI

Quality adjustment in the Irish CPI (PDF 751KB)

[1] Classification of Individual Consumption According to Purpose (COICOP)