Since introducing the Conus Trend series for regional jobs data we have had frequent queries as to why our data appears so different to that provided for the Cairns region by the Queensland Government Statisticians Office (their report for Cairns Sept data is available here). The simple answer is that the Government Statistician’s produce their reports based on a simple 12 month moving average of the available (unadjusted) ABS data. That is, they simply add the last 12 months of data together and divide by 12. The Conus Trend on the other hand uses a complex statistical model (called X-12-ARIMA) which is widely used by statistical providers around the world. By doing so we are seasonally adjusting the raw ABS data and then smoothing that to create a less volatile Trend series. We also make some adjustments to the resulting regional data so that it sums to a total which is consistent with the ABS Trend data for Queensland.
We believe that this gives a more nuanced picture of the state of the labour market in the regions. However, we would always caution that the resulting Trend series (and moving average series) are derived from what are bound to be fairly small data sets. In particular sub-sets of that data (for example data related to just females, or just youth) are necessarily very small date sets and need to be treated very cautiously.
To compare, the table below outlines the Queensland Statistician’s data for Sept (and the changes from a year ago) with the same numbers for the Conus Trend data for Cairns in Sept.
|Queensland Statistician’s||Conus Trend|
As we see, although the numbers for Sept 14 are quite similar for both measures, there are significant differences when it comes to the changes over the past year. The simple reason for that is that the 12 month moving averages used by the Government Statisticians are placing equal weight, by way of the moving average in Sept 2013, to some high employment and participation figures in the latter part of 2012 as they are to the data over the past few months. The Conus Trend is looking at the more recent movements to determine the trend levels and therefore the impact of that data in late 2012 will have had less of an impact on the Sept 2013 Trend; hence the difference in the annual changes across the two measures.
Neither measure is “right” or “wrong”; they are merely different ways of looking at the same data. As we have made clear before, we much prefer the Trend.