6466.0.55.001 - Information paper: Experimental Data in the Pensioner and Beneficiary Living Cost Index, Feb 2013  
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 18/02/2013  First Issue
   Page tools: Print Print Page Print all pages in this productPrint All

APPENDIX EXPENDITURE CLASS ANALYSIS




There are currently 85 expenditure classes in the PBLCI of which 10 make use of detailed expenditure data relevant to PBLCI households accounting for 29% of the total expenditure by pensioner and other government beneficiary households.

Of the remaining 75 expenditure classes, 52 exhibit statistically significant differences in spending patterns and 24 of these are also showing large differences in price movements.

Twenty expenditure classes exhibited no significant difference in spending patterns, while an analysis of spending patterns was not possible for three expenditure classes.

Table 1 below provides a summary.

A1 Comparison of Spending Patterns within Expenditure Classes

Expenditure Classes
Number
PBLCI Expenditure Weight %

Total PBLCI/CPI Expenditure Classes
85
100
Weights within expenditure class derived from detailed PBLCI household expenditure data
10
29
Weights within expenditure class NOT derived from detailed PBLCI household expenditure data; of which
75
71
Show statistically significant differences in spending patterns; and
52
55
Large differences in price movements(a)
24
27
Small differences in price movements
16
17
Limited price analysis possible(b)
12
11
No statistically significant differences in spending patterns
20
14
Analysis of spending patterns not possible(c)
3
2

(a) The difference in price movements was at least 0.5 percentage points over the 5 year period March 2006 to March 2011
(b) There was only a limited concordance between groupings used for the analysis of spending patterns, and CPI elementary aggregates
(c) No grouping of items below expenditure class level was possible using HES data


The methodology employed for the analysis of spending patterns was as follows;
  • Within CPI expenditure classes, commodities were partitioned into sub-classes. These sub-classes were designed specifically to highlight differences in spending patterns, and in general they did not coincide exactly with CPI elementary aggregates.
  • 2009-10 HES data was used to estimate the expenditure proportions for each group calculated separately for pensioner and other government beneficiary households and the households in the main CPI.
  • Retail receipt information was used, where available, to create sub-classes of commodities.
  • Pairwise t-tests using standard ABS methodologies were used to identify statistically significant differences in the expenditure proportions between pensioner and other government beneficiary households and the households in the main CPI.
  • A difference between the two in spending in an EC is considered to be statistically significant if a pairwise t-test has a p-value less than 0.05 (95% confidence) on at least one major category of items in the EC.

The methodology employed for the analysis of price movements was as follows;
  • A concordance was established between CPI elementary aggregates and the sub-classes described above.
  • Where the concordance was sufficiently good, elementary aggregate-level expenditure weights were estimated separately for pensioner and other government beneficiary households and the households in the main CPI from 2009-10 HES data.
  • Elementary aggregate weights were used to weight elementary aggregate-level price indexes from the CPI to produce price indexes for expenditure classes. This was carried out separately for pensioner and other government beneficiary households and the households in the main CPI.
  • This enabled a comparison of the price movements over the period March 2006 to March 2011 based on the weights for pensioner and other government beneficiary households and based on the weights for the general CPI population.
  • While the difference in the price movements due specifically to differences in the elementary aggregate weights is only an estimate of the difference in price movements actually experienced, it can be used to determine whether improvements in the quality of the index can be realised.

Table 2 below lists the 21 expenditure classes for which statistically significant differences in spending patterns translated into large differences in price movements experienced by pensioner and other government beneficiary households compared with the households in the main CPI(footnote 1) . For each expenditure class the table shows the influence of the expenditure class in both the PBLCI and the CPI as measured by its expenditure weight.

A2 Table 2: Expenditure Classes Showing Statistically Significant Spending Patterns and Price Movements

Expenditure Class
PBLCI Expenditure Weight (%)
CPI Expenditure Weight (%)

Fish and other seafood
0.56
0.41
Lamb and goat
0.47
0.26
Poultry
0.69
0.49
Food additives and condiments
0.42
0.30
Fruit
2.35
1.60
Jams, honey and spreads
0.22
0.14
Garments for men
1.11
0.74
Garments for women
2.14
1.47
Maintenance and repair of the dwelling
2.48
2.05
Glassware, tableware and household utensils
0.44
0.43
Hairdressing and personal grooming services
0.80
0.90
Telecommunications equipment and services
3.75
2.93
Audio, visual and computing equipment
1.27
1.56
Audio, visual and computing media and services
0.91
0.98
Other recreational, sporting and cultural services
0.98
1.09
Other services in respect of motor vehicles
1.11
1.35
Cleaning and maintenance products
0.42
0.29
Other non-durable household articles
1.85
1.46
Other household services
0.97
0.69
Newspapers, magazines and stationery
0.99
0.68
Tertiary education
0.30
1.40
Total
24.20
21.20



Aggregate indexes were constructed to represent the 21 expenditure classes in the table above, where statistically significant spending patterns resulted in differences in price movement of more than 0.5% over the period from March 2006 to March 2011. The indexes are shown in Figure A3 below. The index based on the classes in their existing form is labelled the 'PBLCI 21 classes aggregated' and the index constructed using the more detailed HES data to derive weights within the expenditure classes is labelled 'Experimental PBLCI 21 classes aggregated'. The comparison of the price movements revealed that in 16 of the classes, the movement was greater (more positive) in the experimental classes and this has translated into an index level that is around 0.7 points higher than the existing. The 21 classes examined represent slightly less than 25% (in expenditure terms) of all classes in the PBLCI and Figure A3 shows the impact that this differences in the 21 classes would have had on the all-groups index level (naming conventions are maintained). The impact on the all-groups index is minimal over the period considered.

A3 Aggregating Expenditure Class Indexes and the Impact on the All-groups Index, Index level comparisons
Graph: A3 Aggregating Expenditure Class Indexes and the Impact on the All-groups Index, Index level comparisons


1 For three of the 24 expenditure classes, it was not possible to complete a price comparison for the full five year period. <back