When I was at college, one day a sales rep from Wall’s came to give a little presentation. Tempted by free ice-cream I went to hear the talk. Most of what he said fell through and lost in memory during the years except one small fact: apparently in Europe people living in the colder countries consume more ice-creams! I never quite figured out why that would be the case. It’s a bit like none of us understood why our fingers wrinkle during a bath (Well some scientists ‘figured it out’ recently). It’s all real, but it’s also bizarre.
Last weekend while it was snowing heavily, I didn’t eat any ice-cream. But I did switch on the heater for longer time. Nice weather for utilities providers I thought. Capgemini Business Analytics team has a track record of working with utilities companies to help them forecast utilities demand. One important indicator is always weather conditions. This is quite intuitive. But I wonder if there are other industries which might be affected by weather too? I decided to take a particularly closer look at the retail sector.
On the internet, it appears that links have been drawn between poor retail sales and bad weather. And there are 2 general hypothesises: 1) Cold/rainy weather drives down retail sales 2) Cold/rainy weather drives up online retail sales
Both seem to make sense. On a nice warm sunny day, people are probably more likely to go out and shop around whereas when it’s cold and miserable, people may prefer to stay at home and shop online. So is any of the above hypothesises true?
To find out, I built several little statistical models. Thanks to Office for National Statistics and Met Office, I was able to collect some essential data. The key data I used was average monthly temperature, rainfall volume and non-food retail sales data since 1994 and internet sales data (this was only collected since 2010 by ONS). I also included ‘year’ variable to account for inflation trend and ‘month of the year’ variable to account for traditional sales peak such as after-Christmas sale. Key ‘inflation’ and ‘monthly’ characters can be illustrated in the charts below:
So what have my model results shown? 1) The warmer it is, the more sales non-food retailer can expect (Shown on chart below) 2) However, the relationship does not hold between rainfall volume and sales
Given the very limited online sales data, I found: 3) There is no significant relationship between temperature and online retail sales 4) And rainfall volume has no bearing on online retail sales
There are a couple of points to note: firstly, monthly rainfall volume might not be the best representation of how miserable the weather has been (e.g. if it only rained during the week and all sunny during weekend); secondly, there was only about 3 years worth of online sales retail sales data from ONS and small sample size means last 2 results need further validation.
Alright, so if you can tolerate the two little data issues mentioned above, then the myths between weather and shopping have been solved! People in Britain do not fear rain if they had to go out shopping but they don’t like freezing condition... Do you agree?