Substitution issues

Substitution issues

An ideal price index must make some provision for changing purchasing habits and expenditure levels as people switch to new and different products. The average shopping basket and what people choose to spend their money on today is very different from that of fifty years ago. An index also has to have a method for dealing with items that cease to be sold. Inevitably, the approaches taken to deal with these issues can have an impact on the resulting index – often lowering it.

In the UK, the ONS reviews the items on the shopping list every year. It then has one overlap period where it monitors both the old and new list and then “chain links” the two datasets together with statistical wizardry. In the US the review takes place every two years (it used to be every ten) and in Germany every five years.

That process seems sensible on the face of it. However it creates a few issues. Say the price of something goes up a lot, so much that many people choose not to buy it. People have indeed suffered inflation and the new prices are higher than before. Under the ONS system of annual changes, these items will quickly stop being checked and instead are substituted with cheaper items potentially of inferior quality that are not necessarily rising in price.

It is difficult to determine the effects of this. However, if you do look at the constant items in the index (or more specifically some of the food items where ONS publishes the backdata ) and then compare them with the overall RPI food index, you see clear evidence of the increases of the average individual items being greater than the increase in the apparent total index, which adopts the substitutions.

Specific food prices

Source: ONS

In other countries the index suffers less from this effect as components of the inflation are adjusted less frequently. It is difficult to say what is an optimum review period, but a formal review every three-five years would be a good compromise between consistency and the need to reflect changing purchasing habit.

Quality adjustments

Some items checked in indices have short life cycles and are regularly superseded by improved or different versions. This is particularly the case for technology items, such as computers, cameras and mobile phones. How this is dealt with can cause issues for the index. To some extent, the problem in the UK at least is overcome by ensuring the items being surveyed are revised annually – see above. However, improvements can occur during the year and older spec models may even be discontinued.

In the 1960s the US adopted a process called “hedonic pricing”. It involved creating a mathematical model for the prices of items dependent on their features*. Therefore, if a product needed to be substituted for a higher spec and more expensive one, a calculation would be done about how much the higher spec one would have cost at the beginning of the year. This aimed to deduce how much of the change in prices were related to the higher spec features and how much, if any, was a real price increase. They then used it extensively for most white goods (e.g. kitchen appliances), technology items and clothing/footwear.

Many other countries have followed suit in the last decade and the UK for example uses it for phones, computers and cameras.
This is a fine idea in theory, but the simple models can rarely portray the intricacies of pricing. Furthermore, the approach usually ends up in estimating that the price of such items has declined, when the absolute price paid for, say a new phone, by the man in the street may have actually increased, as the models offered for sale have new features for which they cannot avoid paying.

To give you an idea of the impact of such measures on some of the components that go into inflation indices, let’s consider what has happened to the price of some of the items being adjusted this way in the UK. Between 2005 and 2012, CPI went up by 23 per cent across all goods. However, the price of digital cameras had fallen by 83 per cent and that of laptops by 65 per cent. This implied that the average person buying these products was paying massively less than they were in 2005.

Unfortunately, there was no objective measure of the average price paid for a digital camera or laptop to enable us to verify this method. However, as an alternative, you can look at the average price of all products in the Which? magazine review tables from 2005 and compare it to an average price from a similar review in 2012. This gives a fairly representative sample of the products being bought in both periods by real people for the purpose of taking pictures and using a computer. It focuses on the prices people are really paying and ignores any of their features and the clever mathematical models.


Sources: Which? reports** and ONS

The results show that indeed the price paid for both a digital camera and a laptop had fallen since 2005. However, the reduction was not anything like that being assumed in the CPI inflation index on the basis of modelling. This analysis is not a precise measure of average prices of these products in 2005 and 2012, but it does show that government modelling was probably causing at least a doubling of the price reduction effect. Overall this is contributing to a reduction in both of the UK’s main inflation indices. The total impact is not that much, probably around 0.1 per cent or even less***, as the number of items covered by this method is small. Having said that, if the published CPI rate is 1.5 per cent, it would have been 1.6 per cent without this adjustment.

People have argued that in the US, where this technique is much more widely used, it has reduced overall inflation estimates by as much as a third. Clearly, it is a problem trying to determine how best to account for quality changes in products. It would appear though that using mathematical models may not be the most accurate way to do it. However while this practice remains, it continues to help governments make inflation rates look lower than they actually are.

* This takes the form of a regression model, where items have a base price and fixed amounts are then added for each feature that the product has.
** “Digital Cameras: Which?” June 2005 (n = 10 cameras), “Which?”, December 2012 (n = 28 cameras). “Laptops: Which?” April 2005 (n = 13 laptops), “Which?”, August 2012 (n = 25 laptops). Predicted price calculated by taking the Which? price and multiplying the CPI index change from 2005 to 2012 in categories: 09.1.2 Photographic, cinematographic and optical equipment and 09.1.3 Data processing equipment.
*** It is difficult to be precise over the exact effect from the published data and without knowing the unpublished details of the hedonic modelling. However, assuming it is only being applied to certain goods such as PCs, laptops, digital cameras and PAYG mobile phones, its net effect on the published UK prices indices will be to lower them by an amount of around 0.05%. (These items contribute only around 1–2% of CPI/RPI index weights and the hedonic modelling effect in recent years has been to reduce prices by around 10% pa, some of which is a genuine reduction, as was noted by analysis of digital cameras and laptops in the text.)