3. Covec's estimate of z

3.1 Overview of Covec's Methodology

3.2 NZIER Review and Covec Response

3.2.1 Technical Questions

3.2.2 Theoretical Questions

Covec Limited was engaged by MED in 2003 to prepare estimates of advertising revenue growth for TV and radio broadcasters.² In April 2004, we were asked by MED to review the Covec report.

3.1 Overview of Covec's Methodology

1. As advertising accounts for the vast majority of TV and radio revenue, Covec estimated z by forecasting annual changes in advertising revenue.
2. Covec chose population as the best "predictor" of future advertising revenue. The basis for this was that as the population grows, the potential market for any advertisement also increases. This increases the price of any given advertising timeslot, and hence lifts advertising revenue.
3. Covec then derived a statistical relationship between advertising revenue and population for both TV and radio.
4. Using Statistics New Zealand's population projections, Covec used this statistical relationship to forecast future advertising revenue to 2030.

Covec's results indicated that:

• Radio advertising revenue will rise from $203 million to $296 million between 2002 and 2030. This is equivalent to compound annual growth of 1.4% over this period. Over the period from 1990 to 2030, this translates to annual growth of 1.95%.
• TV advertising revenue will rise from $572 million in 2002 to $782 between 2002 and 2030 million in 2030. This is equivalent to compound annual growth of 1.1%. Over the period from 1990 to 2030, this translates to annual growth of 2.02%.

3.2 NZIER Review and Covec Response

NZIER reviewed Covec's paper³ and made a number of comments. The majority of these were quickly addressed by Covec.⁴ Our comments fell into two broad categories: technical and theoretical questions.

3.2.1 Technical Questions

A lack of technical detail in the original report meant it was difficult for us to comment on the technical characteristics of Covec's statistical analysis.

Covec supplied more detailed information, including diagnostics in response to this observation. NZIER was then able to confirm that all diagnostics were normal.

3.2.2 Theoretical Questions

We questioned whether population growth adequately accounted for all the drivers of advertising revenue. We suggested nominal GDP as a possible alternative.

Covec explained that since population drives audience size and ratings drive advertising prices then population drives advertising revenue. They also argued against nominal GDP as there is some circularity in the relationship with advertising, and it is difficult to obtain an accurate regional disaggregation.

While we agree with Covec that there is a relationship between population and advertising revenue, we are less dismissive of nominal GDP as a potential driver. However, as will become apparent in the next section, the point is moot as we agree that population provides a better basis for forecasting advertising revenue.⁵

We noted that technological change and productivity growth had been ignored. However, we noted that while including these was possible, it would increase the uncertainty around the forecasts. Covec indicated that they had explored this question and that:

• Insufficient information about historical technological and productivity improvements exist on which to form a basis for a forecast.
• In the long term, ongoing productivity gains would cause exponential growth in profit margins. This is not a reasonable assumption.

We agree with these arguments.

We initially thought it important to distinguish between the volume of advertising available and its price. The reason for this was that at different times in the past different factors have been more important. In particular, in the late 1980s there was relatively high general inflation, and in the early 1990s the number of television stations in New Zealand increased from two to three. As Covec pointed out though, we are interested in forecasting the value of revenue, which diminishes the need to decompose it into price and quantity changes. Furthermore, the data itself suggested that there was no need to make this distinction.