Review of NZIER report: The impact of the proposed Emissions Trading Scheme on new Zealand’s economy, April 2008

I have been asked to consider the following:

  1. Appropriateness of the methodology and the ORANI model in particular.
  2. Validity of conclusions in relation to the modelling results.
  3. Role and validity of key assumptions.
  4. Gaps in the analysis.
  5. How any of the above might undermine the report’s conclusions.

My comments below are based on reading the report, an exchange of emails with NZIER and a meeting with NZIER.

1. Appropriateness of methodology

A general equilibrium model is a very appropriate tool for analysing the costs of policies aimed at reducing greenhouse gas emissions.  Such policies are likely to have diffuse and far-reaching economy-wide effects.  Thus partial equilibrium techniques such as traditional cost-benefit analysis would be inadequate.  Models of the ORANI type have a long and credible history of use, especially in Australia, such as by the Australian Productivity Commission.  A similar model is used in New Zealand by BERL.

ORANI type models are not unusual from a general equilibrium theoretic perspective. Their special feature is that equations are expressed in logarithmic differential form, making them linear, and thus avoiding the need for complex nonlinear solution algorithms.  The NZIER solution algorithm incorporates a technique for dealing with linearisation errors.

2. Validity conclusions from results

The report’s conclusions follow logically from the modelling results.  The main issue though, is what is driving the modelling results, in particular the finding that the ‘NZ Pays’ scenario has lower national cost than the ‘ETS’ scenario? 

3. Role and validity of the key assumptions

Assumptions are taken to include both the macroeconomic closure assumptions dealing with scenario specification, and the equally important embedded model assumptions.

With respect to macroeconomic closure, none of the assumptions appear to be unreasonable when taken on their own, in particular sets of circumstances.  The issue, however, is more subtle – what are reasonable assumptions to use in particular situations, how do they interact with the embedded assumptions, and how crucial are they to the results obtained?

To try to answer these questions I consider the report’s results for 2025, though some of the discussion will also apply to the results for 2012 and 2015.

In quantitative terms the results are generally of the same order of magnitude as those presented in Infometrics’ research for the Emissions Trading Group.  Differences will be caused by macroeconomic closure assumptions and a whole array of differences (some very minor) around the parameterisation of the model’s production structures, demand functions, industry aggregation, energy substitution possibilities and so on. 

The surprise in the report is that the NZ Pays scenario has a lower national cost than the ETS scenario, whether measured in terms of GDP or private consumption.  While all of the factors mentioned above will contribute to the size of the difference, the direction of the difference would seem to be driven primarily by four main features of the model:

  1. The capital closure assumption – the ETS produces such a strong negative effect on rates of return that aggregate investment falls (relative to BAU), leading to a smaller capital stock and thus lower GDP. 
    Arguing that aggregate investment would fall under an ETS is not necessarily implausible,[ Indeed our own modelling for the New Zealand Business Roundtable looked at such a scenario.] but neither is it certain.  That the ETS scheme results in pressure on profitability in some industries is understandable, but the ETS is about changing relative prices, not lowering profitability.  Indeed some industries see a potential increase in profitability?  Examples (from the report) include fishing, tourism and some manufacturing.  To understand why total investment declines we need to consider the points below.
  1. Agriculture – agricultural exports are extremely sensitive to what happens to rates of return and to costs in general.  Indeed NZIER confirm that a subsidy to agricultural raises GDP – quite apart from any carbon pricing issue.  The loss of profitability in agriculture is a key determinant of the overall reduction in investment.  No land is left idle, but some does shift to less productive use under an ETS.
  1. Taxes – taxes are not explicitly modelled.  This means for example that investment is more affected by the ETS, which lowers rates of return, than by the NZ Pays scenario where taxes are (implicitly) increased – which should affect the user cost of capital.  The difference between how a carbon tax affects private consumption and how income tax affects private consumption is also not explicitly modelled.  Thus the analysis appears to ignore the effects of changes in different types of taxes on allocative efficiency.
  1. Closure with regard to the external balance – Payment for the purchase of international emission units is treated as an increase in net factor payments offshore.  To maintain the current account balance (which is fixed at the BAU ratio to GDP) there is a reduction in imports in the ETS scenario.  In the NZ Pays scenario there is a reduction in imports and an increase in exports.  In effect this is similar to what happens in the Infometrics modelling.
    On the capital account of the balance of payments there is also a reduction in a foreign liability.  Given that, as an accounting identity, the change in the capital account must equal the change in the current account, and that the change in the latter is zero (by the closure assumption) there must be an offsetting change in the capital account.  This takes the form of an increase in foreign investment in New Zealand (which is modelled as a net reduction in capital outflows), which is an increase in foreign liabilities.
    In the ETS scenario the larger contraction of the capital stock means that less foreign investment in New Zealand by foreigners is required than in the NZ Pays scenario.  Looked at in reverse, less foreign investment is forthcoming because of the reduction in the return to capital.  Thus the capital closure assumption is tied up with the external balance closure assumption. 

4. Gaps in the analysis

There are no gaps in the analysis in the sense of issues that need to be addressed, but there are gaps in the explanation needed to understand the results of the analysis – as evidenced by the discussion of the points above.  Having said that, however, there is a limit to the kind of detail that one can incorporate in a public report, and I have no complaints about the willingness of NZIER staff to answer my questions.

5. Robustness of the report’s conclusions

Some sensitivity testing of the results is described in the report, and the general conclusions with respect to those tests are robust.  What was not tested (or at least not publically released) is the robustness of (2025) results to changes in the capital closure assumption, to the changes in the sensitivity of agricultural production and exports to prices, and to changes in different types of tax.  I understand that changing the capital closure assumption has been analysed and that the relativity of the results – namely a greater welfare loss under the ETS than under NZ Pays – is unchanged, although quantitatively diminished.[NZIER make the point that changing the capital closure assumption does not fit well with the closure assumption regarding the balance of payments.]  Changing the agricultural price elasticities is presumably not too arduous a task, but modelling different tax rates requires extensions to the model. 

In summary the NZIER analysis presents a useful contribution to economic analysis of the options for meeting New Zealand’s international emissions reduction obligations.  Once one  understands what has gone into the model, one can understand what has come out – which is not true of all models.  What goes into a model is mix of things we know something about and things we know very little about.  Over time the latter category hopefully shrinks, but many assumptions and parameter values issues will continue to be disputed.  They need to be addressed by sensitivity analysis within the model framework and by debate outside the model framework.

As mentioned above, the quantitative results from the modelling are comparable to those obtained by Infometrics, suggesting that running both models with identical closure assumptions and agricultural demand/supply elasticities would lead to similar results.  Such analysis would be usefully complemented by more detailed research into the sensitivity of investment to pre-tax rates of return and of the sensitivity of agricultural production to prices.

Adolf Stroombergen
Infometrics
19 May 2008