By Kevin Connolly & Mairi Spowage
Since 2017, the Consumer Price Index (CPI) including owner occupiers’ housing costs (CPIH), has been the lead indicator of inflation in the UK. An extension of the CPI, the CPIH gives a more accurate picture of inflation as occupiers housing costs, which amount to a large proportion of overall household spending, are included.
Similar to other measures of inflation, CPIH has generally only been published at a national level, however it would be advantageous to produce and publish estimates at a regional level. The development of regional CPIH would allow for differences in prices to be tracked and could be used as another variable to measure the progress of the levelling-up agenda, something we are considering in current research for the Economic Statistics Centre of Excellence (ESCoE).
The objective of our work is to provide updated regional CPIH estimates in a similar manner used for national values – following closely the methodology outlined in Dawber and Smith (2017). This is an extension of previous work as we now report CPIH rates up to the end of 2020 and investigate different data sources for regional expenditure weight calculations.
Method
The methodology for calculating regional CPIH follows closely, with some adaption, the national calculation, outlined in Figure 1.
Figure 1. Schematic of CPIH national methodology
Fundamentally, the calculation of CPIH involves aggregating the prices of items within a shopping basket (of goods and services), using a series of weights, to estimate inflation. Weights are used as not every item within a basket is consumed equally. As we spend more on some items than others, these items should have a greater influence on inflation rates. For example, you would expect a 10% increase in petrol to have a higher impact than that for the same relative price change in tea. For regions we follow a similar method outlined in Figure 2.
Several challenges and limitations have been identified when applying the national CPIH methodology for regions, including:
- Small sample size in the price quote database
- Improper use of stratum weights
- Volatility of international classification of individual consumption by purpose (COICOP) class weights
The COICOP weights are the focus of this paper. Through discussion with the Office for National Statistics (ONS) prices team, the view was taken that much of the volatility in the weights found in previous work could be explained by the source data used. We therefore take a different approach by investigating further the Living Costs and Food Survey (LCFS) data available from the Secure Research Service (SRS), looking at unweighted estimates and using the regional Household Final Consumption Expenditure (HFCE) for weighted estimates.
Results and discussion
Using the HFCE database to update COICOP weights we were able to produce updated regional CPIH and inflation estimates, shown in Figures 3 and 4 respectively.
One noticeable issue with the regional CPIH and inflation estimates, particularly from 2014 onward, is the sizeable difference in some indices between December and January (Scotland in 2020) for example. This is assumed to be caused by the chaining of indices. These ‘jumps’ occur in the national CPIH index but are much more subdued than the regional ones, indicating that the issue may be in the price/item quote sampling size. With the regionalisation of national data, on average, the sample for each item is a 12th of that used in the national calculation.
There are a number of options here to deal with issues of small classification sample size when regionalising the price quote database. The first is to investigate options for increasing the sample size of these classes across the regions in the price quote collection. The second, is to investigate options for alternative data sources for price quotes.
If we increased the sample size to contain the same level data as the national CPIH estimate the collection would increase by a factor of 12, which is infeasible due to the associated cost. The increase could instead be varied across regions.
We carry out some analysis of the data and set a minimum number of observations of 10 per COICOP classification per region. To achieve this goal, using the most recent year of data, there must be an increase in sample size across 11 of the 12 regions. To achieve this 10 minimum standard the aggregate price quote database size would need to increase by 62%. If the minimum target was set at five observations the database would only increase by 10%, whereas a 215% increase is required for a 20 observation minimum. In addition, this is not a universal increase across all regions, rather a targeted approach.
Directions for future research
In this paper, we use the methodology outlined in Dawber and Smith (2017) to update the Consumer Price Indices including owner occupiers housing cost (CPIH) for the 12 NUTS-1 regions of the UK.
Previous work focuses on the price quote strata and small area estimates whereas in this paper we investigate using a different set of source data from which to calculate the expenditure weights.
Similar to previous work, this paper finds somewhat unreliable results. In short, the adjustments have not led to a significant improvement in the quality of estimates produced. Using the HFCE estimates produced by the ONS does seem more promising than previous work, although chaining indices seems problematic for regional price estimates and should be investigated further. through increasing sample size and exploring other data sources.
Read the full ESCoE Discussion Paper here.
Kevin Connolly is a Chancellor’s Fellow, Department of Economics, University of Strathclyde.
Mairi Spowage is Principal Knowledge Exchange Fellow and Acting Director of the Fraser of Allander Institute, Department of Economics, University of Strathclyde.
ESCoE blogs are published to further debate. Any views expressed are solely those of the author(s) and so cannot be taken to represent those of the ESCoE, its partner institutions or the Office for National Statistics.
