Matt Nolan

The issue of data

/in , /by

I will give people a little more time to write comments on this post on economists and predictions before I discuss it – let me say that I’m very happy with the two comments (from HH and Kimble) so far.

There is a related issue that I think needs a bit of attention before we start discussing these things – data. Statistics New Zealand does an excellent job getting together data and putting together clear, transparent, and useful, data series. However, even with all the hard work they put into the process there will always be some issues with the data, namely:

  • Not all the survey responses arrive, or arrive at the same time – implying that there is constant revisions to the data,
  • Changes in methods mean that there is constant revisions to the data
  • Changes in methods implies that some data series for the same thing (eg private consumption over time) are incompatible,
  • Seasonal adjustment and trend estimates can be subject to an “end point bias” – which implies that the figure we have at the time is different to the “true” figure that will be available in the future,
  • Quality adjustment is a subjective process – but the adjustments are not always transparent,
  • In a country like NZ sufficiently disaggregated data cannot be released,
  • In a small country the “timeliness” of surveys can be poor (GDP is out three months after the event for example),
  • The length of data series – for many types of statistics there just isn’t much data!
  • Even if the definitions of what the data does is transparent – the data will often be misused by people interpreting it.

What does all this mean – well, even though Statistics NZ does the best job in getting data that is possible, and even though economists might make the best use of the data that they can, it takes a long time (several years) until it is clear (in a quantifiable sense) what has actually happened.