I’m giving up on AFG

My long term readers will be aware that I have followed the loan issuance data from AFG since I started blogging. Over time I have noticed quite a few problems with their reporting and data, most notably a large revision in LVR calculations that went completely unmentioned by AFG themselves. Over the last few months I have been questioning whether it is worth continuing tracking their data because of these anomalies and the fact that their overall reporting seems to be out of step with the broader market direction.

Back in early september I noted that AFG had reported a large surge in mortgage issuance:

August mortgage sales surged to their highest volume in over eighteen months according to AFG, Australia’s largest mortgage broker. AFG processed over $2.7 billion of mortgages in August – the highest volume figure since March 2010. Refinancing accounted for 38.2% of all mortgages as property owners took advantage of a highly competitive lending environment.

While August figures are typically stronger than previous months, these figures trend well above seasonal expectations. There was a sharp increase in fixed rate home loans as lenders heavily discounted their rates. Fixed home loans comprised 9.4% of all loans processed compared to 7.9% in July – the highest level since December 2010.

We now know however that there was no surge in housing credit and in fact the housing market melt continued. Given this latest data I have finally decided to give up reporting AFG data as in my opinion is no longer is providing any value to MB readers.

Below is their report for October, it is the last one I will be presenting.

AFG October


  1. I disagree – the data will still stand up, you have your time frames wrong.

    AFG present loans approved, data not really captured by the official data until the lenders report, which is on a slightly different time frame.

    Most loans approved in August will be reported in September which is when they are generally funded, although there is some statistical abberation due to 60 day and 90 day settlements etc.

    You have to be patient, over a long period their trends have proven to be correct, although slightly different time frames are evident.

    Here is my crude graph on the lending figures for AFG – http://www.brisbanebusinessfinance.com/images/Home_Buyers_Graph.pdf

    You will note that column 8 reflects the AFG January 2011 data, which isn’t born out in the ABS figures until Feb/Mar 2011 on memory.

    That isn’t to say the market isn’t generally down, it is, but it isn’t yet out as many here believe.


    • While I haven’t looked for some time, another unusual thing I noticed was that while house prices have very clearly been falling, AFG has only ever shown steadily increasing average prices for their share of the market.
      Are AFG somehow capturing the small segment of the market where prices continue to rise? It certainly doesn’t accord with the RBA’s data, nor with the broad situation anecdotally.
      AFG appears to show the overall situation as being not too bad – strange then that one in every four mortgage brokers have shut up shop in the past year or so.

      • Lefty – the loan sizes contain almost 40% refinanced loans, so that is skewed considerably – I just don’t think that you can go by that as an indication for house prices. refi’s are traditionally higher loans that FTB’s.

    • +1

      The more I look at the AFG data, the more I can see that the trick is to completely ignore the press release.

      I also agree that the last few months has shown a bit of an improvement. I will wait until the end of the year to see if that is a notable trend, or just a statistical blip.

      But I disagree with one suggestion – while FHBs are definitely doing better than 6 months ago, most of the strength in these numbers is from refinancing. Take that out, and the market is looking very weak.

  2. i am following Adelaide, and the number of loans in the AFG data was down 60% from the peak (in 2007), earlier this year, while the ABS data only showed a decline of 40% for the total. With the latest jump, the AFG is now back at -40%, it just catched up with th ABS numbers. It seems so far AFG is just regaining the market share that it lost during the GFC, at least for Adelaide. ABS and AFG data have deviated significantly in the last 2 years.

    • It’s quite true that market deviation will occur, but again you have your time frams wrong. The ABS data will catch up with the AFG data, not the other way around.

      • I’ll try posting a graph later on to show what i mean. Did you try plotting the ABS and AFG data side by side since 2007 for Brisbane?

        • No I haven’t done that, there is no AFG data available for Brisbane, although they do have a QLD section.

          It’s a data capture difference, that is all, but I would be very interested in your graph. I’ll be very happy to admit that I am wrong if you or someone else has evidence, but from my own graphs I can see that the data is 4 to 8 weeks ahead of the ABS data, which is a bonus for me, I don’t criticise the data for that.

          As mentioned in the OP, AFG did have a very embarrassing software glitch after a massive upgrade that caused the reported LVR’s to be incorrect for some months, which they fixed retrospectively, but should have explained correctly. I don’t think that they were aware that people followed their data until then, so I think we should put that behind us and move on.

          Look forward to your graph.

          • Here is the South Australian graph: it compares AFG vs ABS (total number of loans, original, % change from May 07, 3-month moving average)


            As you can see, in 2009 the two numbers diverged quite significantly and the AFG data did not capture the FHOG “boom”.

            Now, after a Feb/Jan dip, which was not reflected in the ABS numbers, the AFG total seems to be coming back to match the ABS level, however it is not clear if it will lead and by how much.

            Clearly there was a jump in AFG data in SA in the last few months, however it is not clear if it will be reflected in the ABS (it could be just an increase in the market share or increased activity in the AFG specific market)


        • MB – here is my FHB graph – http://www.brisbanebusinessfinance.com/images/FHB_Graph.pdf

          My apologies in advance for not being able to tag the Y axis in a monthly time series correctly. To get you bearings the three highest columns are Feb/Mar/April 2009 and the far right column is September 2011.

          It’s hard to argue that the trend doesn’t reflect the exact same trend on the ABS data, although I don’t have a graph on the monthly FHB numbers – do you have one?

          The FHB numbers are now as high if not higher than any numbers we have seen except during the high stimulous period.

          Don’t take that as a “game set and match” prediction of another boom, I just can’t see that happening, it’s just evidence that the market has more life than many bloggers give it credit for.

          • MB – well the stats you pointed to don’t give me FHB alone, only composite totals.

            I have been following the FHB’s because they are the most important section of buyers. If they stop buying the market can only hold up for a short time, so they are the most indicative sector for future trends.

            However I do have a graph of recent months for all AFG home buyers using some calculations to subtract loans for upgrades etc.

            You can view it here – http://www.brisbanebusinessfinance.com/images/Home_Buyers_Graph.pdf

            Column 8 is January 2011. If you look on the ABS graph that you referenced – the ABS figures bottomed in March 2011, not in January like the AFG data did.

            So if my thesis is correct, lending on the ABS data should be falling until September and then pick up in November.

            We shall see in due course. Some variation may occur though, there is always a little fuzziness in the timing of data.


          • MB – I see what you have done. Generally I would argue that the AFG trend still is mirrored by the ABS but still a couple of months difference in the time frame.

            I think you will see a lot more unreliability in the SA stats thn other states because AFG have a smaller presence in that state. WA, Qld, Vic, and NSW all have much greater volumes of transactions, which give those states greater reliability.

            The SA figures are less than half of the other states. As a matter of history, AFG started in WA, so they are strong there, and later got quite a hold in Qld, and also spread to Vic and NSW although not quite with a much percentage penetration in the two larger states.

            Yes I could download the ABS spreadsheet and overlay it when I have time. A little busy today though.

            Thanks for sharing your info with me.

          • Hi Peter, I think you need to look further back in the FHB data, pre-2007. There was a lull in late 2007 (probably due to affordability) where the FHB numbers dropped below their trend. So compared to their trend they are still well below, especially if you account for population growth.

            Your data may reflect the increased market-share of AFG as well, as you mentioned they have expanded.

            FHB numbers are still in a post-stimulus lull and there are not enough potential FHBs out there to bring number back to normal for about another 6-12 months (unless we induce demand again, but that will still result eventually in lower growth due to low FHB numbers). The ABS numbers align very closely with my calculations as the the “normal” level of FHB activity (popn 22mil, around 80% home ownership at some point in one’s life, 80 years life expectancy, approx 1.7 people per FHB transaction = 129,000 FHB per year), this is almost exactly the average over the last four years. That equates to around 10,000 FHBs per month and the latest ABS is still only 7 to 8 thousand.

            One last thing, FHB activity leads house prices by about 6 months, so it’s unlikely that this will improve prices for some time yet even if the numbers bounce.

          • Miss P – Yep I’d agree with that. On memory FHB’s trend at about 24% so at around 17% they are still below trend, and any real estate bounce must have them back at or at least near trend.

            I reiterate, that although I expect prices to soften some more, it’s just that I consider that the softness is at the moment a little less pronounced than many believe.

            But given the many variables in the global economy, any scenario could play out.

          • I wouldn’t look at the proportion of fhb to all as it fluctuates wildly, the total number is better as this only grows slowly with population as a trend.

            The long term average proportion is around 20 per cent, I think, but it’s not a good indicator of the market strength.

      • Peter, as you may remember from the Credit Crunch forums, I have a fondness for 12-month moving averages of raw data. Using that approach I don’t see any consistent time lag between ABS and AFG, in terms of marking trend changes. Possibly because the trend changes aren’t sudden reversals.

        In short, ABS and AFG broadly agree that the last 3 trend changes occurred Jan 2009 (up due to FHOG Boost and rate cuts), October 2009 (down due to FHOG Boost reduced and rate rises) and June 2011 (up for reasons unknown to me). The latest trend change is the least convincing of the three, if one excludes refinancing.

        Although DE has seemingly identified some issues with the AFG data, I still think it’s useful in predicting ABS data. And obviously very useful that it’s published a month before the ABS figures.

        • Yes I remember. I have only used raw monthly data. Happy to provide the excel spreadsheet if you would like it. You could probably improve it for me.

          I agree that the 3 month moving averages would probably give us closer results.

          I think that the data is good, but not completely understood, which rather surprises me in some cases.


        • Merovingian, Peter.

          Thanks for the feedback.. I will take another look at the trend data and make another assessment.

          I still do have some serious concerns about the data because of its lack of seasonal, trend and compositional adjustments, and as I said in the post I have previously found errors in the data.

          However I will give AFG the benefit of the doubt and try and find some time to build a few charts comparing their data ( with moving averages ) to the ABS data.

          I agree it is useful to find a datasource that is “ahead” of the ABS.

          Thanks Again


          • Thanks DE – please take a look at my reply to DT below, that should explain the time lags for you. The recent differences that you have agonised over, have yet to surface in the ABS data.

            I’m not sure how you would adjust for seasonal data with any accuracy, given that contracts are weekend driven, and not necessarily related to working days in the month.


  3. Personally I believe we need an apoloitical objective body that does this ‘stuff’ on a regular basis and is not the ABS.

    No disrespect to the ABS but we need timely accurate data – I would have no issue if this body was national,abnd they sold their data to all buyers and released aggregate data to the public in a bi-monthly cycle.

    We need a truly single objective point of view for what is one of the biggest purchases anyone makes (generally) in this game of life.


    • Isn’t that what the ABS is supposed to do?

      Honestly, if government doesn’t have sound, objective data… how is it supposed to govern?

    • Why?
      You only buy a house you can afford where mortgage costs are maybe 20% of your after tax income.

      Stats are for people who want to overleverage.

    • Close, they are approved loans. As you indicate they differ from settlements both in time, and very slightly in volume as some loans do not settle.

      The exception to that was January in Qld when MANY approved loans did not settle due to the flooding, and nor were those uncompleted contracts contested due to a direction by the Qld Law Society.

      • Peter

        I humbly think you are incorrect. Unless something drastic has changed the AFG data is from their mortgage application aggregation system and all from brokers. Theoretically all applications should eligible for approval. Also many approved loans dont settle because brokers put in multiple applications to a number of lenders trying to get the best deal. This would severely distort the figures.

        Happy to be proven incorrect but I think the disclosed size of the total AFG book indicates that about 40% of applications do not settle

        • Sorry deep T but you have been misled. Brokers do not put in multiple applications, that ruins any chance of getting any of the applications through with the credit scoring system. I believe you are a banker, so I’m not sure why you haven’t taken the credit scoring into account.

          Furthermore the system doesn’t count applications, it counts approved loans, so there is generally minimum variation in the figures, except for the January Qld figures that I mentioned, when a huge number of approvals did not complete for obvious reasons.

          I have just spoken with David Michie from AFG on this matter, and he has confirmed the above.

          It is important to have due regard for the above information, as it completely changes the timing of the data, and it must be taken into account when comparing the results with ABS data.

          Does that clear it up for you?

          • Not really. I’m just pointing out that for whatever reason AFG do not settle anywhere near $2-2.5bn in loans per month as shown in the data.

            However, it does not mean that the data is not useful

  4. Ah well you have me there, I don’t have access to the accounts.

    The corporate profile does say that they process more than $2B finance each month, and I would take that to be settled loans, but there is no way of verifying that. Data on approvals that don’t complete is meaningless.

    I just can’t see the incompletes being any more than 10% – if in fact it is that high. Time doesn’t permist shopping to very lender in town, and the credit scoring makes it counter productive.

    Anyway, let’s leave it at that.