How do you solve a problem like the housing market?
I hear this question quite often: how is the housing market doing? Let’s try to answer that.
First, define what precise information you want to know: whether the cost of buying (or renting) a property is going up, or down, and by how much, in a particular area (say Brighton).
Second, find a source of data. There are plenty of housing market indices and reports, many of which are national, which is no good. They also differ by what they measure: prices of completed transactions, or prices agreed (but not completed: many sales are either renegotiated, or break down), or prices advertised (asking prices). We are looking for an early indicator of trends, so let’s go for asking prices. Considering we want the actual raw, low-level data, for asking prices, in and around Brighton, the simple choice is to look at rightmove.co.uk. They provide listings of asking prices, by partial postcode (BN1, BN2 and BN3 will do).
Third, identify the limitations of the data. Asking prices are optimistic by nature, so will be over-valued, but that’s ok because we’re looking for a trend, not an absolute number. Rightmove.co.uk only covers 90% the whole housing market (or so they claim), but does it matter? No. The data will also contain duplicates (properties advertised by several estate agents), and that is a (small) problem. The data will also ‘group’ some properties, for example with a single advert for a whole new development. Again, a small problem. Finally, the data may contain ‘junk’: properties already sold, or which were never on sale (estate agents are a funny lot). Ideally you should try and quantify these ‘problems’, and estimate whether they will affect your data significantly. In our case, it’s unlikely, particularly because we will…
Fourth, aggregate the data into useful, reliable information. There is a little bit of anecdotal value in the price of a particular property - but very little. The real value, what we want, is to see the overall picture. If my lovely neighbour at no 35 sold her flat for x thousands pounds more than she bought it - is it due to partly, only, or despite her refurbishing skills? The right thing to do is to regroup, summarise, merge the low-level data into high-level information: the average cost of housing. We can do that for each partial postcode (BN1, BN2 and BN3). We should do that using a median - a more meaningful metric than a simple mean. By summarising the data into information, we a/ produce something more useful, more open to interpretation, and b/ reduce the noise. Averaging the cost of properties does remove most, if not all, of the problems we had with the data.
By regularly repeating this measure, say every 2 weeks, or once a month, you can form a historical picture of the housing market. Which is what we set out to do.
Wax on, left hand. Wax off, right hand
And how to you solve a problem like Brandwatch? The same, really.
- Decide what you want to measure (mentions of a brand or keyword, sentiment, topics of discussions, etc)
- Find the sources - good, relevant ones
- Identify data weaknesses - but do something about them: no need to crawl the whole web (but get a good sample), remove spam, remove duplicates, etc
- Aggregate the data into valuable, reliable information
So what about the housing market in Brighton?
Since January, ‘for sale’ asking prices are down 4%, and stock (properties for sale, not sold yet) keeps increasing, +30% now. It’s a different picture for the rental market: asking prices are up by 20%, but stock has increased by more than 50%… A lot of optimistic buy-to-lets out there (dare I say their end is nigh?). And the picture is consistent over BN1, BN2 and BN3 (a good sign).
Finally, you should turn the information into action. Buying doesn’t sound like a good idea now. Keep renting if you are already, and look for landlords which have been there for a while, i.e. who don’t have an expensive mortgage to cover. I wouldn’t rent a property bought in the last 5 years (check the land registry online).