You may have noticed that we are asking for forecasts on a range of monthly economic data releases for the US. These sorts of questions might seem a bit daunting because the method of forecasting is different (asking for a number rather than a probability) and because few people will have ever attempted to forecast economic data before.
Given that these questions will be repeating monthly though, I would suggest that there is plenty of incentive for people to get interested. Some simple excel work should get you forecasting with acceptable accuracy each month and if you manage to build a model that out-predicts the crowd, you should have a reliable monthly source of points. It’s worth investing some time and effort.
Today I want to quickly introduce the US Retail Sales number which will be released tomorrow morning at 08:30 New York time. Let’s start by addressing what exactly this number is.
It is the month-on-month percentage change in the value of sales in the US domestic retail and food services industries. The value part means that the number reflects the amount of dollars spent rather than the amount of goods sold. In other words, if prices go up and the amount of goods sold stays the same, this measure of retail sales will go up.
The number is also adjusted for seasonality so, for instance, the expected Christmas retail surge is adjusted for. If December 2016 showed a 0.0% change monthly change in sales, this means that the Christmas rush was no larger or smaller than might be expected given the level of sales in November. It doesn’t mean that there was no Christmas rush.
I don’t have the space here to run through all the details of how to forecast retail sales numbers, perhaps more another time, but here are a few hints as to where to start.
This is the US Census Bureau main page devoted to the retail sales index. The latest monthly release, along with a good deal of detail, can be found here. You can download historic data here. For a bit of background you can listen to webinars run by the US Census Bureau.
The general trend in retail sales growth is something you might have to work out for yourself but I have a few pieces of advice for you in making an estimate each month.
The first is to pay attention to the accuracy (or lack thereof) in the underlying data. The monthly official releases show the estimates along with an error range. Some detail around these errors is given on the final page and it’s worth studying. Given that the almanis market is based around forecasting ranges into which the number might fall, an appreciation of the inherent unpredictability of the number is as important as guessing the most likely outcome.
There are a couple of items within the retail sales numbers which tend to be more predictable that the others. Sales in gasoline stations, for instance, make up 7.5% of total sales and follow the price of gasoline pretty closely (chart below). This makes sense because most of what people pay for in these places is gasoline and they tend to consume more-or-less the same amount of it irrespective of the price. Weekly gas prices are available from the US Energy Information Agency.
Sales of vehicles by auto dealers makes up around 19% of overall retail sales and can be very volatile from month-to-month. Here again we can get a bit of foresight into what to expect each month because the number of cars sold is published by motorintelligence.com. As the chart below illustrates, big monthly swings in vehicle sales can make a real impact on overall US retail sales.
One final thing, don’t look too closely at the various measures of US consumer confidence. It turns out that the actual value of money spent by consumers has little to do with how confident they claim to be when surveyed. At least not on a month-to-month basis.