Wednesday, August 6, 2014

Economic Modelling: a device for bamboozling the public

A column from veteran economics correspondent Ross Gittins at the SMH today includes some cynical comments on economic modelling by consultants: How the renewable energy target affects the cost of living

Disregarding the headline, Gittins takes issue with conflicting economic modelling by two leading economic consultants ACIL Allen and Deloitte Access. His conclusion:
Why have two leading economic consultants reached such opposing conclusions? Perhaps because Deloitte’s modelling was commissioned by the Chamber of Commerce and Industry, the Business Council and the Minerals Council.
Deloitte doesn’t conceal that its modelling is in reply to ACIL Allen’s. Would it surprise you if the fossil fuel industry wanted to see the renewable energy target abolished and was alarmed to know that modelling commissioned by the review had demolished the argument that continuing the target would add to people’s electricity bills? Now the review will be able to pick whichever modelling results it prefers.
How did Deloitte reach such different results? By feeding different assumptions into its model. It seems to have assumed the cost of wind farms won’t fall over time (which it probably will), whereas the price of gas for gas-fired generators won’t rise much (which it already has).
Regrettably, economic modelling has degenerated into a device for bamboozling the public.
Deloitte has also provided economic modelling for the Aquis EIS. I had intended a thorough review of the economic and social chapters of the EIS prior to closure of submissions but procrastinated too long with weekend diversions. However some things noted from the economics chapter include such as this commentary and supporting data:
The recent decline in mining and resource activity across Queensland has also been felt in Far North Queensland where unemployment rates have climbed up to 7.0%, compared to the rest of the State at 6.1%  (see Table 13-3). Workforce participation for this region is also somewhat lower, at 59.6% as opposed to 62.2% for the Queensland level.

The narrative and table appear rather inconsistent here. Unemployment in FNQ only recently climbed to 7% with the resources slowdown? The table appears to refer to 2011 when resources were going gangbusters rather than anything recent? I can't reconcile the numbers for either unemployment and participation rates with anything from any relevant ABS monthly series. The employment figures are presented as FTE (full time equivalent) so perhaps there has been some adjustment for this somehow or the data is related to the 2011 census?

There is modelling on tax revenues but I can't find any reference to what gaming tax rate has been assumed, although they have provided a discount rate for NPV. I previously posted on different gaming tax rates in Queensland with Reef Casino currently at a concessional rate. There is also reference to employment relative to a seemingly undefined counterfactual.

How much validity then should we attribute to any modelling? I would contend that an appendix with appropriate supporting detail and assumptions on modelling may at least be appropriate, as is also suggested by the terms of reference for the EIS. However all that appears to be provided is the following in references: "Deloitte Access Economics (2014). Aquis Resort internal memo regarding economic profile and impacts."

The Initial Advice Statement from Aquis last year included output from KPMG modelling with employment projections widely spruiked in the PR blitz. Any attempt though for either Aquis or the Coordinator-General to release the actual KPMG report was futile. Interesting aspect there is that the number of direct construction workers seems to have shrunk by several thousand with the revised proposal submitted in the EIS.

Renowned model sceptic Richard Dennis from the Australia Institute also posted again recently on economic modelling which includes discussion of input-output and CGE models: Economic models often biased by vested interests:
The only thing that can be said with certainty is that at least one of the modellers is 100 per cent wrong.
The Deloitte modelling for Aquis is CGE.


  1. Having had a quick look it would appear that the FTE numbers have been derived from the Working Population Profile (table W03c) from the 2011 Census. Use hours worked (mid of range), multiply by the number employed for those hours, divide by 35 hours and sum the you something very like what you have in the table.
    Unemployment and PR also derived from the relevant 2011 Census data.

  2. Thanks. I suspected some census link to 2011. Hardly "recent"!

  3. True...and odd to use such old data when what they're after is fairly straightforward to get.