Dozens of studies have been done on the sales effects of advertising, using aggregate time-series sales data for brands, coupled with ad spending. Meta-analyses of these studies has arrived at average advertising elasticity figures of around 0.1 to 0.2. That means, a 10% increase in advertising spend results, on average, in a 1 to 2% sales increase, in the short term. Short term here means from the date the ad airs, to about 4-6 weeks later.

But a new study finds a much lower ad elasticity figure, only around 0.03 – not 0.1 to 0.2. The authors use a massive dataset from Nielsen of over 200 US brands including weekly sales and advertising spend. The authors Shapiro, Hitsch and Tuchman suggest their lower figure is more accurate than what comes from meta-analyses because of the ‘file drawer problem’ – only published work gets used in meta-analyses, and there is a bias against publishing studies that show null or counter-intuitive results such as sales going down during advertised periods.

The implication of this piece of work is that on average, the short-term effect of advertising on sales is extremely low, almost zero. For established packaged goods brands, at least. That’s not to say it doesn’t work – we know it does. But it doesn’t work to noticeably boost sales in the short term. It works via reminding consumers and refreshing memory. More detail about how advertising does work is in the books How Brands Grow and How Brands Grow 2 as well as Building Distinctive Brand Assets.

The managerial implication is, brand managers should not expect their advertising campaigns to increase sales in the short-term. But they also need to realise that a large part of the the reason some brands are big is their investment, over years, in brand-building via advertising. And they won’t stay that way unless they keep doing it.

Note, the study makes a distinction between short term and long term effects, but in reality it only measures the effect of advertising over a fairly short time-span. To measure a long-term effect they set a decay parameter for brand advertising whereby an ad in one week has a ‘carryover’ effect the week after, and the week after (this is known as the ‘adstock’ model). For example they used a decay parameter of up to 0.9, so one week after the ad is aired, its sales effect is assumed to be 0.9 of the first week’s effect, then (0.9×0.9 = .81) 0.8 in the second week, then (0.9×0.9×0.9) = 0.73 et cetera.

Note, this study relates to established consumer goods brands. It did not consider advertising contexts such as price-oriented retailer advertising.