RPData Index – Charted

The monthly RP Data – Rismark House Price Index was released this morning for May with the raw data showing a fall of 0.5% after last month’s fall of 0.1% was revised to -0.3%. In seasonal adjusted terms the fall in May was 0.3% after last months 0.3% was revised to a fall of 0.4%.

As the chart above shows there is no hiding from the fall in house prices both in Nominal terms but equally in “real” or inflation adjusted terms.

Now we have modelled the relationship between the quarterly moves in the ABS house price data series and the total monthly value of housing finance ex-refinancing and we see a 0.97 correlation since the mid 1980’s.

So add this data in with the above correlation and today’s RBA credit aggregates and you get an outlook for house prices that suggests further weakness ahead.

Latest posts by _EcoRon_ (see all)


  1. No surprises in these numbers.

    My favourite quote from the press release –

    “The softening in Australian home values is delivering a valuation dividend with Australia’s dwelling price-to-disposable income ratio falling to 4.2 times, which is its lowest level since June 2003 according to Rismark’s analysis.”

    What is a valuation dividend? It seems to imply that falling prices are good because homes become more affordable. That’s all good and well, but buying into a falling market is probably not a good investment decision, when your net yields are <3% and your leveraging is detroying your equity at a rapid rate.

    • You should know to ignore that “ratio” Cameron – it borders on meaningless.

      And they still have four PhD’s on staff but can’t draw a chart worth a damn.

      • Umm..RPData does not make any claims about the number of PhDs they have 🙂
        The 4 PhDs are with Bullhawk/Rismark. I presume they are locked up in a cage – I never hear their names mentioned in the MSM.

      • Mav, the data comes from RPData, the index is calculated by Rismark.

        The “price to income ratio” is calculated by Rismark.

        The chart that shows this must be done by someone at Rismark.

        Its not even undergraduate level quality composition – but I’ve talked about this before – just because you have a PhD or 4 doesn’t mean you know how to represent data correctly or fairly.

        • Exactly. However, I know leading software developers of data visualization solutions who have no idea how to present a graph. Go figure.

    • Yes, I noted that phrase as well. It took me a moment to even figure out what they were trying to spin there.

      My favourite line that is being thrown around by everyone at the moment is that house prices “retraced by -1.2 per

      We’re not walking back along a known path, we’re falling into an unknown future.

  2. And if one could justify the oddity of applying a bit of TA to these series, one can see a prominent lower high on all of them at about the Jan 11 time mark, particularly in the blue line (Real NSA). Though the two HPI series are almost both double tops.

    I’m probably pushing it a bit, but it is fun to look for pivots that were not as evident in the Jan 08 downturn.


  3. “Now we have modelled the relationship between the quarterly moves in the ABS house price data series and the total monthly value of housing finance ex-refinancing and we see a 0.97 correlation since the mid 1980’s.”
    This is a huge claim. Please provide precise details of the variables which you have found to be so correlated and the resultant regression line equation.

    • Deus Forex Machina

      I have just re-run the correlation and regression myself on the data we have and I am comfortable that the result is correct.

      SW if you are sharp enough to ask the question then I’m sure that you appreciate that the very first thing your professor tells you in stats 101 is that others must be able to rely on your data.

      Although we are not perfect we hold this tenet dear to our hearts and we always try to adhere to this rule.


      • Alex Heyworth

        This is not a huge claim at all, given that the comparison is between two statistics that are essentially different ways of measuring the same thing. It would be a huge surprise if they did not have a high correlation.

        • Deus Forex Machina


          a bit like the correlation between night following day…or is it days follows night…

          aaarggh to hard for a simple bloke like me 🙂

          • The_Mainlander

            Thank you again I have learnt more in one post than years of analytics spin doctors at work! 😉

      • I understand that they should have a high correlation (so I’m not being critical here) – but I think SW’s asking for the method used. I’ve been trying to find correlations with various data myself recently, so I’m interested in knowing what you did here: E.g.

        * what abs series does the quarterly prices come from? I can only find back to 2002 (but I’m not very adept at navigating abs.gov.au)

        * where does the housing finance ex-refinancing come from? Is this multiple RBA series combined?

        * how did you convert monthly to quarterly? Did you just subtract to get the difference?

        * which variables did you actually correlate? are you looking at changes in one variable to the absolute of another?

        * any lag?


    • Do RP data / Rismark ever provide “precise details of the variables” when they produce similar analysis? It’s a bit rich that you demand such info from MacroBusiness but don’t seem to hold the private sector data suppliers to such high levels of scrutiny.

      How about you declare to all of us what your interest is Suzi (or should I say Sarah P / Chris Joye?)

      • The relevant data at issue is public data. It should be a very simple matter to state the variables which it is claimed to be 0.97 correlated.
        It’s quite a different matter to expect commercial operations to reveal private data.
        The tone of your response indicates some sensitivity to the issue and confirms the appropriateness of my request.

        • No suzie, I just get frustrated when people like you demand such info in a self-righteous manner. Yet you fail to hold the data providers or MSM to such high standards.

          How about you ask Rismark for the data underpinning their spurious dwelling price-to-income ratio. You know, the one that says Australian dwelling prices are only 4.2 times household disposable incomes (suggesting average household disposable incomes in excess of $100k).

      • “…Suzi (or should I say Sarah P / Chris Joye?)”

        Whomever he/she is, he/she seemed to have agued with the confidence of a person with a full day’s advance knowledge of the rather unexpected manner in which the RBA announced that they are changing the rules* and said the rest of the world was wrong and that good asset quality along with assured central bank support is all that need worry a bank treasurer, rating agencies, non-domestic lenders and domestic lenders, nothing to see here, we are one happy family, taxpayer funds are bottomless bank support funds, it’s different here.

        * by changing the rules, I mean Jumping the Shark. All downhill from here.

    • Suzi, I wish to validate your request as reasonable…

      However, i think the data relied upon by Data Sword is actually implicit in the very sentence of Data Sword’s that you quoted:

      Specifically: “the ABS house price data series” and the “(ABS – implied??) total monthly value of housing finance ex-refi” (my edits).

      Further validation from a critic’s point of view could be demanded from Data Sword, as he could provide the ABS data table numbers which were used – possible, Data Sword?

      However, Suzi, i’m curious about your final request: to provide “the regression line equation”.

      My question to you regarding this are:

      1) do you mean a linear (ie. “line”) equation? If so, Data Sword certainly would not have used a linear regression, and thus cannot provide a linear equation. Perhaps it’s fair to ask Data Sword what regression software package was used, and this might give some insight into what regression method was used; though, regardless of the package, the resultant regression equation (ie. y = f(x)) would probably just be a high-order polynomial.

      If I have misunderstood you request, it is possible that others might have also – could you be more specific?


      • Personally I don’t see the need for the regression equation – it sounds like a simple linear regression, i.e. using the correl() function in excel.

        But, understanding the regression equation would tell you if there are any other terms used in the analysis which might increase the correlation, such as seasonal adjustments. Again, it doesn’t sound like there are to me, and it wouldn’t bother me either way if the regression equation was provided or not – I’ll build it myself.

      • No, I wasn’t assuming a linear regression but even a nonlinear relation will still only contain the two variables and 2 or 3 constants which I would have thought could be stated.
        I raise the issue because I don’t see that level of correlation. If we take the ABS housing finance table 1 column M (total commitments less refinancing by value) and compare with the ABS house 8 cap city indices we can see that the figures are all over the place.
        For example, the 3 year period 3/1991 to 3/1994 saw the monthly finance commitments (less refi) rise by 158% whilst house prices rose only 9% over that whole period.
        Another example shows that for the 2 year period 3/2006 to 3/2008 monthly housing finance (less refi) actually fell by 14% whilst house prices rose by 24% over the same period. I haven’t included investment loans but I don’t believe that significantly changes the picture.

        • Well, if you’re using EXACTLY the same info, then you’ve got a fair “please explain”, in my view.

          …ARE you using exactly the same info?

          Data Sword, could you please provide your exact references?

          Also, Suzi, are you using monthly or quarterly data (more resolute data, such as monthly, will have more volatility, and, thus, a lower correlation coefficient – broader-period data (as long as it is not TOO broad) will produce a better correlation, as volatility is smoothed out. Data Sword used monthly HPI data with quarterly ABS credit data.

          (Data Sword: on that point, above…someone might have a valid point in questioning the validity of the correlation, given that you’ve mixed your apples, so to speak – monthly and quarterly….)

          Another point: i don’t think Data Sword used the 8-cities index; I think he used to Nationalised Indices in all cases? (am I wrong??)

          Though I’m tending to trust what analysis Data Sword has done, I’ll admit to being “surprised” that he got a 0.97 correlation coefficient….that is simply amazing, and virtually unheard of….

          …So, IMHO, some cynicism is definitely warranted.

          • I’m not surprised. He made Schoolboy Econometrics Error #1: correlating two trending variables that are probably both “integrated”. This is known as “spurious correlation” and is taught in any second or third-year econometrics course, and probably any decent first-year statistics course.

            At the very least he should detrend first, eg by differencing. Google spurious regressions trending. Second hit is http://www.autobox.com/spur14.html which is a nice simple explanation.

  4. Apologies for knit-picking, but is the labeling meant to read June 05= 100 for all series, or have I got it mixed up..

    • Your graph shows aggregate housing assets and M3. There’s a very strong correlation there. I’ve seen it similarly expressed as the correlation between median house price and per capita M3. That correlation has been strong as far back as the RBA’s M3 data goes ie 1959.
      However, the correlation of aggregate housing assets with total credit or housing credit is very much weaker.

  5. It is interesting to watch the monthly revisions from RP Data / Rismark, especially as most are downwards (except for the extreme negative results, which generally get revised upwards towards zero). For example, a few months Sydney was revised down from 0.8% (which was hailed by property “journalists”) to -0.3%.

    The lazy journalists who regurgitate the press release often quote last month’s numbers, without checking the revisions first.

  6. ceteris paribus

    Confused. Clarification please on the above article please. Is housing price correlated 0.97% with housing credit levels or the change in housing credit levels?

    • 0.97 is the Pearson Correlation coeffcient.

      Roughly speaking, is gives a indication of how “in step” two variables are with eachother (but keep in mind that correlation does not necessitate causation, but can be part of a case to argue so).

      However, i am not sure if Data Sword is using it as the squared or non-squared version…

      …Data Sword: is the 0.97 “R” or “R-squared”??

      • Although at 0.97 it hardly makes a difference, I will jump in an guess R, based on excel’s correl(). R-squared is still 0.94.

  7. From what we have modelled, best regression can be obtained from the number of housing finance approvals excluding refinancing. Value of housing finance approvals work, but not as well.


  8. To the people who say of course to the correlation above of 0.97 has anyone considered the possibility of foreign investment into the equation? The fact that the correlation is so high possibly shows to me how much the housing market is debt driven rather than cash from external sources. I’m quite sure I read somewhere that this correlation factor has been rising over the years as houses become more and more debt financed. So it isn’t “night and day” rather it shows how dependent housing has become on new debt issuance.

    The big banks probably know this though and have/will start relaxing requirements again.

  9. ceteris paribus

    If the absolute value of housing finance is the correlate, house prices should still be going up.

    Please enlighten.

    • Deus Forex Machina

      Hi hi…we use the flow of housing finance ex-refinancing not the stock of outstanding finance. if we used the stock you would be right about prices – roughly anyway we reckon it will rollover.

      the flow, or demand for housing finance ex-refi has been falling

      we use the trend but the actual number is just as good

  10. CJ seems to be saying ‘New mortgage credit to BUY’

    I disagree… ‘net NEW credit created to BUY’

    The difference between what CJ is asserting and what I’m asserting, is that some credit already exists on some houses so the NEW credit has to extinguish the existing credit.

    If NO NEW credit was created then over time the asset can only transact at the existing credit + existing deposit.

    i.e. Sellers would have to meet buyers at a price equal to the existing deposit base as a maximum aggregate capital price and likely a whole lot less in a functioning economy.

    The debate is… Does the creation of NEW credit money inflate asset prices.

    I would argue YES it is 100% correlated in a credit money system when the credit money IS being directed toward housing asset transactions, and that is exactly what we are looking at with housing credit. Credit for housing purchase.

  11. Guys, Suzi…

    Bit busy here DS will do something seperately later on today

    cheers DS

  12. Hello,
    I am wondering if anyone has access to the RP data median price for every month over the last few years? They only report a moving 12 month windows (currently down 7.5%). But if each moving 12 month window is down, then the 18 month losses may be even more. Without the actual data would be hard to do though.

  13. Susie

    M3 is a very broad definition of all deposit money

    If credit creates deposits (I believe it does), then the RBA D02 M3 Credit aggregates for Investors and Owner captures all the money created against housing as an asset… loosly all housing money created.

    The graph shown is NOT all M3 only housing M3 and housing M3 growth.