Executive summary

The Ministry of Economic Development (MED) commissioned Covec Limited to develop formulae for pricing television and radio spectrum licences. These formulae are based on estimates of broadcasting advertising revenue growth (denoted z) out to 2030. MED later engaged NZIER to review the estimates.

The process of estimating an appropriate value of z incorporated several stages: initial development, peer review and then establishing an agreed estimate.

Covec developed a method for forecasting advertising revenue, the main component of revenue for radio and TV stations, based on population.

PriceWaterhouseCoopers and NZIER then separately reviewed the method. A number of questions were raised in NZIER's report, which were clarified by Covec. These included providing additional detail on the technical characteristics of the equations, and why productivity growth should be ignored. These were satisfactorily resolved.

NZIER then completed some additional analysis to derive estimates of z, with the objective of establishing whether they would differ significantly from those derived by Covec.

We explored a number of options including using nominal GDP as an explanatory variable. The results of our analysis suggested that nominal GDP added additional complexity to the forecasting process without clear benefits. We therefore agreed with Covec that a population-based formula was most appropriate.

Although the basis of our forecasting models was the same some differences in the results remained. After discussing these with Covec we were able to attribute them to

• Rounding of the data.
• The definition of a year (i.e. calendar or March years).
• The precise specification of the data used for forecasting.

In the final step, Covec and NZIER worked together to establish a mutually agreeable estimate of z. Identical data was used to re-estimate both organisations' equations to eliminate the issues listed above. Remaining differences between the estimates can therefore be attributed solely to method differences.

In short, Covec's model is based on a linear relationship between advertising revenue and population, while NZIER has used an approach known as error correction modelling. The choice between these two methods is largely a question of modeller preference.

Covec and NZIER agreed to define the recommended z factors as averages of the values stemming from their respective approaches. Based on averages of these estimates, we therefore recommend the following z values be used for calculating spectrum license renewal fees:

Television: z = 2.60%

Radio: z = 2.43%