Wednesday, February 10, 2010

CPI update

I was attempting to create a 'Make your own CPI' spreadsheet that uses the ABS price indexes for each commodity group and allows you to assign your own weighting.  I was also using the weighting from the German CPI as a comparison (which weights housing as 30% of the basket, while Australia's CPI has housing at 19% of the basket).

But there was a problem.

The price indexes for each commodity group were so completely manipulated by quality adjustment (and any other unknown statistical manipulation that occurs) that you could not dramatically change the final CPI figure.  Using German weightings I was within one percent (of the total change) over a 10 year period.

Have a look at the house purchase index used for the CPI against the ABS' own capital city house price index in the graph below.  The extreme magnitude of the disparity is quite shocking.  The index used for the CPI increased by 36% over the 7 year period, while the capital city median price index increased 92%.
What lesson should we take away from all this?  I for one will never trust an official statistic without first understanding the methodology behind it.


  1. the picture has failed to load for some reason. VERY keen to see this though

  2. So, 3 questions from me:

    Where does the CPI housing purchase price index / rental price index come from?

    Why does the ABS produce 2 similar yet different measures?

    Are the differences all down to "quality" adjustment??

  3. From what I can tell from here, the house price index used for the CPI comes about as follows:
    "A sample of project home builders is approached to obtain prices for a number of specified types and models of project homes. The types of project homes selected are those most commonly constructed in each city. For marketing purposes, many builders provide bonus deals, which can involve upgrades to fittings, extra features or even extra rooms. These bonuses change frequently and due to this price volatility, new homes are priced monthly" Essentially, it is a construction cost index - not sure how much quality adjustment occurs, but it does for rentals (only upwards adjustments are considered due to capital improvements. " quality adjustments are made to the rents collected to take account of the slow deterioration in quality of the dwellings")

    I doesn't tell us much about the real price of housing in the marketplace in which people actually live.

    I've been learning a lot about this. At first I thought there were simply theoretical points about weighting and quality adjustment, but it seems there is more to it.

    But maybe in the end it is a reasonable way estimate the CPI. Land is after all an asset, much like the ownership of a company. We wouldn't factor in the price of BHP shares into the CPI, so include land?

    It does show nicely though that a land price bubble is probably occurring in Australia right now. How can this single asset class escape the plunge in value experienced elsewhere in the economy?