The effects of efficiency dividends

Looking at the ABS stats on retail sales for November, I was struck by this comment:

Due to the smaller sample size and sampling methodology (independent samples for each month of the quarter) in effect from July to October 2008, there is increased volatility in all series. The original and seasonally adjusted series are most affected and, as a result, the original and seasonally adjusted series are considered of limited use for measuring month to month movements. The ABS recommends using the trend series for this analysis.

To meet the Rudd Government's efficiency dividend targets, ABS was forced to reduce sample size. This is one result.

Postscript

This post was a bit too cryptic for Neil.

So called efficiency dividends were introduced as a public sector proxy for competition. The argument runs this way.

Competition breeds efficiency and productivity gains. Government Departments are not exposed to competition and therefore have no incentive to improve productivity. So what we will do is to reduce their core budget each year by a small percentage equivalent to the productivity gains that they should be achieving. Then new activities can be funded as a separate matter.

It sounds so reasonable.

In the case of ABS they had to make some choices in an environment with high staff costs as a percentage of spend and no control over salaries. One thing that they did was to reduce sample size on surveys, thus reducing data collection costs.

The costs of this decision are summarised in the above quote.

As you reduce sample size, sample error increases. In this case to the point that two key parts of the data - the raw data and the seasonally adjusted estimates - could no longer be relied upon.

For those who don't know the term seasonally adjusted, original data is affected by a whole range of things.

Months vary in length, while sales go up and down over the year - higher retail sales at Christmas are an obvious example. Seasonal adjustment attempts to take these factors out of the data so that the real trend can be seen more clearly.

The ABS comment means that reduced sample size destroyed part of the value of a key economic indicator at just the time we needed it. That, I would suggest, is certainly an efficiency dividend. Just a negative one.