CONTENTS

Measuring Advertising Effectiveness:
Expenditure vs.
Gross Rating Point

Voluntary Funding of Commodity Promotion Research Programs

Editor's Notes

Director’s Corner

Next Meeting

Measuring Advertising Effectiveness:
Expenditure vs. Gross Rating Point

by Chanjin Chung and Harry M. Kaiser

In evaluating generic advertising programs, researchers have typically used advertising expenditures as a measure of advertising intensity. This approach mostly assumes there is a constant relationship between each dollar spent on advertising and its impact on sales. Another measure of advertising intensity is gross rating point (GRP), which has been frequently used in the marketing literature. GRP is a product of the reach of the advertisement and the average of its distribution of exposures delivered to a target audience. GRP is a direct measure of physical advertising exposure, while expenditure is an indirect way of measuring consumers' exposure to advertising programs. The purpose of this study is to compare advertising impact and effectiveness using alternative measures of advertising intensity including expenditure and GRPs.

The assumption of a constant impact on sales per dollar expended on advertising can be divided into two assumptions: (1)the cost per exposure is constant, and (2)the relationship between an exposure unit and its impact on sales is constant. If there is a constant cost per exposure (e.g., per GRP), given constant advertising effectiveness per exposure, both expenditure and GRP measures should be equivalent and produce the same evaluation results. However, casual observations suggest the assumption of constant cost per GRP may not hold in any practical applications. First, the per unit cost of GRPs, in general, decreases as GRPs increase due to volume discounting. It is a well known fact in any negotiated business, such as media buying, that large buyers can extract price concessions in the form of discounts. For sellers, it is also true that it is easier to conduct a few transactions with one buyer than several transactions with several buyers. Therefore, the more a buyer is willing to buy, the less per unit cost becomes. Second, the per unit cost of GRPs differs across air-times, target audiences, and regions. For example, costs-per-point for daytime vs. prime time, teens vs. adults, and large vs. small cities are all different. In this case, the assumption of constant cost per exposure does not hold. Furthermore, the assumption of constant advertising effects on sales per exposure is also likely to fail because researchers are dealing with different media products. To maintain the basic assumptions, researchers need to develop some ways to make GRPs from different times, audiences, and markets comparable. Unfortunately few studies have been completed in this area.

This study uses post-buy actual GRPs and corresponding re-cap advertising expenditures as alternative measures of advertising intensity. The two data series are continuous from the first quarter of 1989 to the third quarter of 1998 for the New York State fluid milk market. Fluid milk sales data were obtained from the New York State Department of Agriculture and Markets, and other required data in the evaluation were compiled from various publications.

With these data, we first conduct several correlation tests, and then estimate a simple advertising model to evaluate the effectiveness of the New York City fluid milk advertising programs using two alternative measures of advertising intensity: expenditures and GRPs. We finally compare results from the two measures. Although we expect that the historical series of GRPs and expenditures are highly correlated, the high correlation does not warrant the same evaluation results.

Both parametric and non-parametric tests, on levels and percentage changes, strongly suggest that no association between GRPs and expenditures is highly unlikely. However, this does not necessarily mean these two series are equivalent in terms of magnitude and direction. In particular, if these two series do not move in a parallel fashion, the equal evaluation results are not warranted. We illustrate this graphically in Figures 1 and 2. Figure 1 presents the historical trend of the two advertising intensities in terms of levels, while Figure 2 illustrates the trend in terms of percentage changes. In Figure 1, since the two measures have different scales, we normalized the data with mean values of each measure. The graphical illustrations clearly show that expenditures (deflated by MCI) and GRPs (adjusted) do not move in the same direction. For example, observations 3, 4, 5, 6, 9, 13, 19, 20 and 29 in Figure 1 and observations 3, 4, 5, 13, 20, 31, and 34 in Figure 2 show that the two series move in different directions. Also, in many cases, the differences in percentage changes were quite large.

Figure 1. Advertising Intensities (in a normalized unit)

Figure 2. Advertising Intensities (in a percentage change)

The graphical illustrations indicate that the two measures are likely to produce different results in advertising evaluation. Ordinary least squares (OLS) is used to estimate a simple demand equation, specified in a double-log form for the New York City fluid milk market. Since the demand equation was specified in a double-log form, estimates of parameters in Table 1 can be interpreted as elasticities. In general, key economic and demographic variables had the correct signs, but some were not significantly different from zero.

The estimated own price elasticities of demand ranged from -0.381, in the model using GRPs, to -0.514, in one of the models using advertising expenditures. The estimated income elasticities were almost identical in all three models, but were not statistically different from zero. The relative small magnitude of price and income elasticities in all models is consistent with previous studies of fluid milk demand in New York City reflecting the nature of milk as a staple good. While the two demographic variables and the eating away from home variable had the correct signs and large elasticity values, none were statistically significant from zero. The relative small magnitude of price and income elasticities in all models is consistent with previous studies of fluid milk demand in New York City reflecting the nature of milk as a staple good. While the two demographic variables and the eating away from home variable had the correct signs and large elasticity values, none were statistically significant from zero.

The estimated advertising elasticities were positive and statistically significant in the two models based on expenditures. The elasticities in these two models, 0.071 and 0.075, are also similar to previous studies in this market. Our results indicate that choice of deflator, CPI or MCI, does not yield significantly different advertising elasticities. However, the estimated advertising elasticity for the model using GRPs was not statistically different from zero.

The estimated rate of return in all three models was above 1.0 indicating positive net benefits of advertising. The largest return was found using the expenditure model with MCI as a deflator, which was 11 percent higher than the model using CPI as a deflator. Both expenditure models produced significantly higher estimated rates of return than the GRP model. This suggests that while GRP and expenditure series pass both parameteric and non-parametric correlation tests indicating a high positive correlation, the two measures still have substantial differences in terms of estimated advertising elasticities and rates of return. Thus, the choice of measure for advertising intensity can produce different evaluation results.

Additional research on this issue may enlighten our understanding on the use of alternative measures of advertising intensity in evaluations of advertising effectiveness. Certainly, we need better deflators when researchers decide to use the expenditure measure. Ideally, one would want to have deflators that remove all the differences between GRP and expenditure measures so that advertising evaluation results do not depend on the choice of advertising intensity measure. If this is not practical in the near future, it would be helpful to construct deflators that can account for regional differences in inflation especially in the media market. To maintain consistent and comparable GRP series across historical time periods and different regions and target audiences, we also need better adjustment procedures. A complete procedure should ensure constant cost-per-point as well as constant advertising effect-per-point, particularly when a constant parameter model is used for the advertising evaluation.

Table 1. OLS Estimates and Rate of Return from the New York City Fluid Milk Advertising Programs
	Expenditure MCI	Expenditure CPI	Adjusted GRP
Constant	0.6170	0.6650	1.0603
Price	-0.5141* (-1.77)	-0.5011* (-1.73)	-0.3812 (-1.16)
Income	0.1483	0.1552	0.15.46
Advertising	0.0715* (1.79)	0.0750* (1.91)	0.0444 (0.55)
AGE019	-3.5173 (-0.45)	-3.6296 (-0.49)	-3.2506 (-0.36)
BLACK	3.3346 (0.75)	3.4379 (0.77)	2.9068 (0.58)
EATAWH	-1.0415	-1.0475	-1.1251
Q2	-0.0422* (-3.31)	-0.0425* (-3.29)	-0.0445* (-3.00)
Q3	-0.0537*	-0.0536*	-0.0579*
TREND	-0.0100 (-0.75)	-0.0106 (-0.79)	-0.0095 (-0.63)
R-Square	0.5984	0.6040	0.5547
D.W.	2.122	2.088	2.150
N	3.57	3.18	2.22
*Statistically significant at the 10% level.   Numbers in parenthesis are t-ratios