|
Blog: 26 July 2016 (Unadjusted vs Seasonally adjusted data)
|
Related Topics |
|
Today's blog takes a look at retails sales data, both unadjusted and seasonally adjusted, to get an idea of just how different spending patterns in certain months are, and how different the growth rates are when the data has been seasonally adjusted. And how misunderstanding what the data set is that one is looking at could lead to serious misinterpretation of data.
|
Unadjusted vs Seasonally adjusted data
So what is the difference between unadjusted and seasonally adjusted data? Well unadjusted data is the data as reported or calculated without adjusting the data for any seasonal influences (such as increased spending over holiday periods, greater rainfall in Gauteng during summer months etc), while seasonally adjusted data makes adjustments due to certain seasonal variations.
A good example of where seasonal adjustment is would required is for example spending over the easter holidays, but since easter is a moving holiday and can fall in either March or April, it can have an undesired effect when trying to look at underlying data trends. As easter moving from March in one year to April in the next year could lead to dramatic increases in the month to which easter moved, or rapid declines in the month in which easter is no longer in.
A good example of where seasonal adjustment is would required is for example spending over the easter holidays, but since easter is a moving holiday and can fall in either March or April, it can have an undesired effect when trying to look at underlying data trends. As easter moving from March in one year to April in the next year could lead to dramatic increases in the month to which easter moved, or rapid declines in the month in which easter is no longer in.
For example, in 2015, April was in easter. During these holidays spending at retail stores increases. Thus pushing up April's sales figures. Next year (2016), easter falls in March, thus increased spending in March 2016. The increased spending in March 2016 for easter will not be repeated in April 216, yet in 2015 April sales increased due to easter. Now when comparing year on year sales for example, sales growth from March 2015 to March2016 would show strong increases since easter took place in March 2016. Yet when comparing April 2015 to April 2016 sales there would be a sharp decline since easter took place in April 2015, but not in April 2016. And its these kind of moving holidays and seasonal spending patterns that seasonal adjustment aims to get rid of, in order to provide economists and analysts with a better picture of the underlying data movements free from the impact of seasonal variations.
Below a line graph (blue line) that shows retail sales (in constant prices. I.e. adjusted for inflation) without any seasonal adjustment applied, and line graph (red dashed line) that shows retail sales data after seasonal adjustment has been applied.
As can be seen from the graph above, there is a clear seasonal pattern in the unadjusted data (blue line). Retail sales sees a massive spike every December during the Chirstmas holiday period, followed by a sharp decline in January of each year. Now looking at the unadjusted data on a monthly basis one could easily make the mistake of thinking things are going swimingly when looking at December's growth on November's numbers. Or think the economy is flushing down the drain when comparing January's sales numbers to those of December. And its due to this that seasonal adjustment is applied to data in order to better reflect the actual underlying trend in data.
The graphic below shows the difference in the month on month growth rates for the unadjusted vs seasonally adjusted retail sales data. From this it is clear that the unadjusted data is far more volatile and could lead to serious misinterpretation of the data.
Users and readers should therefore make sure they know which data set they are looking at and what exactly has been done to the various data sets to get them where they are. Is it current or constant prices, is it seasonally adjusted or not, are we looking at Rand values or index levels, is the data annualised or not, are we comparing 3months to previous 3months, or 3months to same 3months of the previous year. There are a thousand different figures that are published and made available. The trick is to cut through the fat and get to the numbers that tells the truest picture of the state of our economy. And that number is usually data that is in constant prices (been adjusted for inflation), and seasonally adjusted (seasonal effects have been removed).
