Saturday 10 November 2012

A Comparison of UK House Prices – November 2012 Update

The last time we looked at UK House Price Indices we ended up with more questions than answers.  I believe there is now a simple answer to why that was the case which I’d like to demonstrate in today’s post.  To help with this we need to first take a step away from the published Rightmove, Nationwide, Halifax, Acadametrics and Land Registry House Price Indices and instead start with an analysis of some very raw data.  This data comes courtesy of the Land Registry and is the database of actual sales in England and Wales for the month of September 2012.  It is a list of 64,173 sales.  If we plot these sales as a histogram we end up with the chart below.


Click to enlarge

A couple of quick observations before we run some analysis on this dataset.  Note the distortions caused by government intervention with the Stamp Duty Land Tax (SDLT) thresholds being clearly visible at £125,000, £250,000 and £500,000.  You can also clearly see the distribution has a very long tail to the right of the peak but is very compressed to the left.

Before we can understand the analysis we also need to understand some basic statistical terms:
  • Arithmetic Mean.  This is more generally called the Mean or Average.  It is simply the sum of a series of values divided by the number of values.
  • Geometric Mean.  This is a little more confusing but can be calculated by taking the log of a series of values, then Averaging these values and then converting back to a base 10 number.  Sounds confusing.   Instead, let’s think about why we might use a Geometric Mean for House Prices.  In simple terms a Geometric Mean tends to dampen the effect of very high or low values which could bias the Average.  In the house price case we have a very long series of high house prices and so this statistic will dampen that affect.
  • Median.  If we list all values in ascending or descending order it is the middle value.
  • Mode.  This is the value that appears most often.

Let’s now analyse this House Price Data using our statistical terms:
  • Number of house price values used is 64,173
  • Minimum house price in the dataset is £9,000
  • Maximum house price in the dataset is £27,925,000
  • Arithmetic Mean is £244,036
  • Geometric Mean is £190,840
  • Median is £184,000
  • The Mode is a little more difficult as the visible government distortion affects this.  Allowing for the distortion it is the £245-£250,000 frequency.  If government distortion was not in place I suggest it would likely be £115-£120,000.

Let’s now look at the five common house price datasets that are readily published which we can use for comparison against this date:
  • The Rightmove House Price Index.  It calculates its house price by simply taking the Arithmetic Mean or Average asking price of properties as they come onto the market.  This means it will be affected by price changes, if the mix of house type changes and if the mix of location changes for houses coming onto the market.  It is not seasonally adjusted and covers properties from England and Wales.  This is not unlike the Average I have calculated above however I use bought prices rather than asking prices.  Asking prices in October were £243,168 which month on month is a rise of 3.5% and year on year is an increase of 1.5%.
  • The Acadametrics House Price Index.  This index uses the Land Registry dataset but in a different way.  It calculates its house price by taking the Arithmetic Mean or Average of bought prices.  It then mix adjusts the data to take a constant proportion of property types, from a constant mix of geographic areas.  It is seasonally adjusted and covers properties from England and Wales.  It covers buyers using both cash and mortgages.  Buying prices in October were £225,954 which month on month is a small increase of 0.1% and year on year is an increase of 2.3%.
  • The Halifax House Price Index.  This index is based on buying prices of houses where loan approvals are agreed by Halifax Bank of Scotland.  It uses hedonic regression to remove type and mix variations thereby measuring the price of a standardised house.  I use the non seasonally adjusted dataset and it covers the complete United Kingdom.  Sales prices in October were £159,818 which month on month is a decrease of 0.4% and year on year is a fall of 2.7%. 
  • The Nationwide House Price Index.  This index is very similar to that of the Halifax except it is based on buying prices of houses where loan approvals are agreed by Nationwide Building Society.  Sales prices in October were £164,153 which month on month is a rise of 0.1% and year on year is a fall of 0.9%. 
  • The Land Registry House Price Index.  This index uses repeat sales regression on houses which have been sold more than once to calculate an increase or decrease.  As it analyses each house and compares the latest buying price to the previous buying price it is by definition mix adjusting its data also.  This is then combined with a Geometric Mean price which was taken in April 2000 to calculate the index.  It is seasonally adjusted and covers properties from England and Wales.  It covers buyers using both cash and mortgages.  Sales prices in September were £162,561 which month on month is a decrease of 0.3% and year on year is an increase of 1.1%. 

It is important to note that all of these price changes show nominal changes.  If we correct for the devaluation of sterling through inflation then we see a very different picture.

To further complicate matters there is then a timing shift between these five indices which we must also consider.  Firstly, a house is placed on the market for the first time (the Rightmove Index).  Secondly, somebody possibly buys the house using a mortgage (the Nationwide and Halifax Index).  Finally, the purchase is registered with the Land Registry (the Land Registry and Academetrics Index).  The best estimate of this timing shift is shown in the chart below which is taken from the paper by Robert Wood entitled A Comparison of UK Residential House Price Indices

The chart below then shows an overlay of all the indices plus the “Actual” Mean, “Actual” Geometric Mean and Median which we calculated from the raw data above incorporating the time shift phenomena.  It uses the Land Registry/Academetrics house prices as the baseline date.  I then average the Nationwide and Halifax house prices in an attempt to increase the accuracy of these datasets as they are very similar and move the data forward by 3 months as that phase occurs 3 months before the Land Registry.  I then also shift the Rightmove house prices forward 6 months as that phase occurs 6 months before the Land Registry.  By doing this we should be able to fairly compare each of the indices with each other as we should be comparing the houses and prices from approximately the same period. 


Click to enlarge
 
So what does all this tell me:
  • The reason I struggled to come to a conclusion in my previous posts on this topic is that I was trying to understand the relationship between all of the indices.  Today’s post shows why I struggled with that and that’s because there is no relationship between the commonly published house price indices.  Each one measures a completely different Average or Standard House. 
  • The Rightmove Index is just about useless.  It doesn’t tell us what is happening to house prices as it only covers asking prices plus it can also be skewed by location and property type mix variations month to month. 
  • If somebody asked me what is the Average House Price in the UK then in my opinion the answer would now be I don’t know.  This is because none of the published house price indices actually tell us this as they all do something with the data they collect before creating their house price index.  If somebody asked me what is the Average House Price in England and Wales then again the published indices are no help.  I would however now be able to state that the Average Price is £244,036 which is what we calculated today from the original data however after stating this I would ask are you sure you don’t want to know the Geometric Mean. ..

I can’t help but feel all of these house price indices are a classic case of “Lies, damned lies and statistics”.

Meanwhile back in the real world I remain out of the housing market as I believe house prices are still overvalued.  As always your opinions/thoughts are welcomed as we move through this house price maze.

As always DYOR.

4 comments:

  1. Hi RIT,

    Great post.

    You may find it interesting to create and compare that histogram from the same months each year from 2006 onward.

    I'd be surprised if it didn't both help you understand the current right move high and also highlight just how few people have been able to take advantage of price falls.

    Cheers

    A1

    ReplyDelete
  2. Hi A1

    Cheers. It's been a bit of a journey but hopefully a beneficial one for readers. It's certainly been enlightening for me.

    You mention reviewing data back to 2006. I've only been able to get raw data going back to February 2012. Do you know of a source going back further that is freely available?

    Cheers
    RIT

    ReplyDelete
  3. So...

    ..the house price statistics are unreliable.

    If so, why are house prices overvalued apart from the reason you can't afford one you would like?

    ReplyDelete
    Replies
    1. Hi Anonymous

      I'm not saying the statistics are unreliable at all. They measure what they are designed to measure. What I'm saying is that each one measures something very different and that there is no relationship between them.

      I believe houses are over valued by looking at historic Price to Earnings Ratio's. The most recent analysis is in one of the links within this post.

      I think you also misunderstand me. I also believe houses are currently Affordable (that analysis is in the same link) and could comfortably afford one that I would like (of course remembering everyone has different requirements - some more modest than others). However, I choose not to buy over valued assets and instead deploy my capital elsewhere for now.

      Cheers
      RIT

      Delete