This recent news clip from the National House Building Council in the UK (NHBC) blaming the highly politically charged drop off in housebuilding to the Beast from the East is rather typical of the way people use the weather – or any other extraneous factor – to explain business performance:
The ‘story’ is then faithfully passed on by the news organisations. Now more than one person has said it, it becomes ‘fact’. It goes back into the organisation. It’s truth. Has anyone quantified this? Possibly. Possibly not.
An independent clothes shop in my local town closed down the other month, giving Brexit as the primary cause. It really does help if the cause has made front page news (note the expression used is the “Beast from the East”, not just “the weather”). It’s Brexit, and not “economic conditions”.
Now don’t get me wrong, I don’t know much about housebuilding or clothes retailing, and this may well be true. But if I do have a gripe, it’s more about the process by which we are able to make such statements. Which is of interest to me because it’s what I do all day. As well as taking the scientific view that says that claims require proof.
What I do know about is modelling weather for other sectors like windscreen replacement, soft drinks, coffee and beer. So, on that basis let’s plunge in.
A not-unpleasing picture of some weather
Everybody LOVES an anecdote, especially from senior management
Interestingly the NHBC Chief Executive does not say, “I asked my team for an analysis of regional weather data year on year and made this considered conclusion”. Instead he says, “I was on site and nothing was happening, the ground was frozen and there was a biting wind.” This happens very often. “I was in my local supermarket and our product was out of stock,” or, “I saw our competitor’s ad last night on TV”. I don’t doubt any of this. It’s a powerful thing. It’s also close to worthless. All that this type of anecdotal evidence can do is help us create hypotheses that we can then test.
“He really said that?” An employee laughs heartily on hearing about another of his boss’s anecdotes
There’s a bunch of stuff going on
In any business there are loads of things going on. In really complicated businesses like housebuilding I suspect there are many factors: raw materials, labour supply, building regulations, capital availability and I don’t know what else. The only way to do this is through building econometric models. That allows us to calculate the actual impact of weather on housebuilding, controlling for all the other variables that affect construction. Think of it as a factor that says for every degree colder it is, you lose so much construction.
The solution to the construction/snow conundrum. Simple, when you think about it.
Weather is multidimensional
How much worse was the weather this year? Well, we need to quantify the weather. Measure the Beast. We have already seen two factors mentioned. The frozen ground and the biting wind. Helpfully we can get statistics on this. Much more easily now that (reluctantly for those who used to sell it), it’s all had to be made open source. (It is however in a terrible format and painful to extract and tabulate. Luckily that’s what we do all day.)
We can look at many things:
- Average temperature (not that useful in modelling, oddly enough)
- Minimum daily temperature (ahhh, now that’s interesting)
- Maximum daily temperature (handy for ice cream and soft drinks, less so here)
- Rainfall (no idea, might be handy, building sites flood I guess)
- Mean windspeed (goes right through you doesn’t it?)
- Number of days of snowfall (snow has a double effect in that people also can’t get to work)
- Days of Thunder (wasn’t that a Tom Cruise film?)
Any many, many more.
We may need to combine these things to create new variables. It may be that it’s not about the wind, it’s about COLD wind. We could, for example, create a ‘biting wind’ variable where the temperature was say below zero and the wind speed above a certain level.
It’s a year-on-year effect
Not only that, we have to establish the impact of the CHANGE year-on-year. We know that house registrations fell by 14% year-on-year for the first quarter, the worst percentage fall since 2012. The question then changes from, “How cold was it?” to “How much COLDER was it than last year?”
Extensive statistical analysis informs us that it’s cold in the winter. The Chief Executive can help us here – he says that exactly (my round-number alarm bell is about to go off…) 30 days were lost in 2018 to frozen ground. But he was less helpful in providing the essential context of how many days are typically lost to frozen ground, so we don’t know how exceptional 2018 is.
Here’s a crazy thought. Let’s spend 2 minutes gathering some weather data. Certainly, February and March had lower minimum temperatures than 2017. There was one really cold week at the end of February. But January was milder. In fact, if we look at average temperature (not shown), there were no days with an average substantially below zero in January 2018.
The weather, year-on-year. Took us 1 minute 51 seconds.
We would probably expect to lose some days to frozen ground in most years. But this effect might have been counterbalanced by the milder January which, curiously, no one mentioned in all of this.
Now we have both parts of the jigsaw. We know much colder was it and our econometric model tells how much temperature (or wind or whatever) affected housebuilding. So that’s it. Isn’t it? Multiply the two together and there we are, all ready for the press release.
The story has to hang together, or we have to say “We don’t know”
The same article also helpfully shows us that this downturn was particularly pronounced in London. Now weather is clearly a regional thing, and regional models can be very helpful here. Production in the North West was actually substantially UP.
The BBC website, ever ready with a helpful graphic
That’s interesting. So, either:
- The weather in the North West was dramatically less bad than in London and weather in London was much worse (year-on-year) than in any other region; Or
- The weather (which we have already been told is the single biggest factor explaining registrations), was dramatically counterbalanced in one region by other (by definition, less important) factors
More Beasts from the East. Mongolian yak head West, leaving chaos, destruction and a trail of missed house-building targets in their wake
The point of this is not to say that it wasn’t the fault of the Beast from the East. It may well have been. The point is that if we make claims, we must have proof, and proof that stands up to scrutiny. Otherwise we risk saying things that aren’t true, these things becoming accepted fact and nobody questioning them. And when we stop questioning our assumptions, our business really is in trouble.