Housepricecrash.co.uk graph is an optical illusion

I assumed people in general would have checked that sort of thing - if you dont you havent much place making analysis of the data!

In any case, whilst you are right that the severity of the graph is highlighted, the bottomline remains the same - current pricing is way above that which one would expect from models based on the long term.

Just because people should check things like that, doesn't mean that they do. Graphicacy and statistical literacy isn't as high as it should be - and not everyone who looks at HPC will be analysing the housing market in minute detail. Plotting the same data in a different way can be revealing, it can also be used to argue a different intention. For example, we could take that same nationwide data and plot the percentage overprice in the actual inflation-adjusted price to the exponential growth trend. (i.e. 100*(actualprice - trendprice)/trendprice)).





Now instead of dwarfing the last peak as presented in the HPC version, the current peak is a fairly similar height to the last one at the moment, and given the shape of the last trough it looks like it could stay flat or indeed get quite a bit higher (after all the previous peaks were quite a bit smaller than the 80s one). I would argue that the (real-price) crash doesn't look so imminent in the second graph as it does in the first, even though both are produced from the same data.

zar wrote:

Just because people should check things like that, doesn't mean that they do. Graphicacy and statistical literacy isn't as high as it should be - and not everyone who looks at HPC will be analysing the housing market in minute detail. Plotting the same data in a different way can be revealing, it can also be used to argue a different intention. For example, we could take that same nationwide data and plot the percentage overprice in the actual inflation-adjusted price to the exponential growth trend. (i.e. 100*(actualprice - trendprice)/trendprice)).





Now instead of dwarfing the last peak as presented in the HPC version, the current peak is a fairly similar height to the last one at the moment, and given the shape of the last trough it looks like it could stay flat or indeed get quite a bit higher (after all the previous peaks were quite a bit smaller than the 80s one). I would argue that the (real-price) crash doesn't look so imminent in the second graph as it does in the first, even though both are produced from the same data.

I only see one graph, should there be something else there?

Or am I being blinded by statistics again?

EDIT: Oh another graph has snuck in since I last looked

talksalot81 wrote:

My concern is that if you are not statistically literate, you shouldnt be trying to discuss it! There are so many people out there spreading misinformation which is not intentional but is done out of ignorance.

As for the graphs - it is not the y-axis shift which is making the difference to the arguement, it is the baseline (i.e. you translate it to be exponential). Without the baseline change, the £50k start pt makes little difference to the look.

I think your interpretation is hopeful. The data itself shows no real evidence. This is of course not to say you are wrong, just it should not be predicted from that chart.

Another point you should note. If we saw the same fall today as we did in the late 80s on the exp plot, prices would not decrease by £40k odds as they did back then, they would drop by more like £70k. So whilst the adjusted chart is much better for analysis, the real effects do need to be considered in terms of the initial chart.

I am pleased to read this thread because it confirms that some people do have an idea of good analytical technique!

Thanks for your comments.

I am a statistics student, so I'm not exactly the average man in the street - and I'm pleased to see the thread too, although its probably not the people who wouldn't be able to analyse the graph who are reading, but never mind!

I was just having a go at playing devils advocate with a different spin on the same data - I'm just starting my statistical career after all, and am still learning the ropes on how to mislead/influence people. I agree the pattern looks like there will be large falls in real terms, and also that the implied time span will be at least 10 years until prices are at their cheapest - if it is an inverted version of the 90s. But none of the previous booms and busts matched the previous one in scale or timeframe so predicting a bigger, longer boom (and a bigger bust?) this time is not a terrible prediction either.

All the exponential model has done is tried to explain what has happened in the past as best it can. It isn't an economic or mathematical model considering household income, interest rates, number of property shows on TV and that kind of thing. And the blue HPC graph - like in the video on the HPC site where it asks what happens next <fall in real prices> vs <stagnation in real prices> - also invites you to assume the pattern will continue... look up a little, down a little, up a bit more, down a bit more, up a lot...

If you are pretty sure you know what will happen next, you are using past statistics to make a prediction, something that many statisticans are reluctant to do. I'm actually pretty bearish on property, although I hestitate to say it as I don't want to invite the usual "oh no not another will there won't there be a HPC thread" comments. I'm as guilty as anyone of looking at these graphs and thinking I know what will happen next. That's the non-statistican in me though - the statistician in me would go "mmm" and "ahh" and say "weell... we don't really like to extrapolate you know..." and then scurry off before any more awkward questions are asked. :whistle:

So returning to the OP... yes the graph may be misleading at a glance, but the information needed to interpret it is all there (y axis labels etc.) and there is nothing really wrong with zooming in on the area of interest if it is appropriate to do so. It is presented in a way that complements the site's views, as are any graphics in the media. The importance of people being able to cope with this is why the Royal Statisical Society campaigns for statistical literacy education in schools as something that all people need in modern day-to-day life.

Being armed with such a graph without knowing how to use it can be dangerous. _pale_

cordial wrote:

As far as this 'statistically illiterate' observer can see, house prices are definitely continuing to crash... upwards...

Yep - good old House prices - always go up - never down!