Measurement Bias in Price Indices: An Application to the UK's RPI

The paper assesses the potential for systematic discrepancies between the measured RPI and an 'ideal' cost-of- living which, the paper argues, should be the target variable for monetary policy (note that the RPI does not claim to be a cost-of-living measure and so this discrepancy does not imply that the RPI is wrong). It identifies four main sources of potential bias. First, product substitution bias which is caused by consumers substituting away from relatively expensive items over time (given that the RPI is based on an annually fixed basket of goods it may overestimate the actual increase in the cost-of-living of consumers who actively substitute). Second, Outlet substitution bias which, in a similar way to product substitution bias, is due to consumers moving away from relatively expensive retail outlets. Third, quality adjustment bias which arises when price rises due simply to improved quality of goods are not fully removed from the RPI. Finally, new goods bias which comes about when new goods, whose price tends to fall, are not introduced into the index. Using the results of detailed studies from the US and Canada and combining these with available data in the UK the paper comes up with a 'guesstimate' range of overall bias in the UK of 0.35% to 0.8% percentage points per annum. (i.e. RPI inflation is 0.35% to 0.8% above cost of living inflation). However, this range does not represent the range of possible outcomes it simply reflects the range of results in the North American studies. In practice, there are a number of factors that these studies have not controlled for and so the possible range of the bias could be much wider than the range defined above.