How Much Advertising Moves Sales

ConversionAssmus, G.; Farley, J. U.; Lehmann, D. R. (1984) · 1984Journal of Marketing Research

Topicsadvertising elasticity·meta-analysis·carryover·replication analysis·econometrics·cpg

Your measurement team hands you a strong advertising number, and a media plan is riding on it. Do you fund the plan on that figure? Pause first. Across established packaged-goods brands, the reported short-term advertising response varies widely: the raw average across 128 models is about 0.2% of sales for each 1% increase in ad spend, and adjusting for how each study was built raises the typical estimate to about 0.7%. A high figure often reflects a model that omitted other sales drivers or mishandled carryover, the effect that keeps working after spend stops. Treat any lone number as a starting point.

With advertising's effect on sales, how the number is measured matters as much as the market it came from.

Across 128 econometric models, the average short-term advertising elasticity is about 0.22: a 10% increase in advertising spending is associated with roughly a 2% increase in short-term sales for mature consumer products. That benchmark confirms the intuition that advertising works. The more important finding is how much the number moves with modeling choices alone. Adjusting for how each study was built raises the typical estimate to about 0.7, and the single largest choice, leaving carryover out, shifts it by about 0.34 on its own, so a number can be high because of the model rather than the market. Knowing that is worth as much as knowing the average.

Data chart

What a model leaves out inflates advertising's measured effect in a single period

A single-period advertising number can look big only because the model left carryover or other sales drivers out.

Action guide

Before you trust an advertising number, ask whether the model counts carryover.If it does not, the figure overstates what advertising did in that period; discount it.
Ask whether the model accounts for other sales drivers like price and seasonality.Leaving them out inflates advertising's apparent effect, so treat a bare-bones model's number as too high.
Use these benchmarks as your reality check.For established, frequently bought packaged goods, the most-cited average is about 0.2% of sales per 1% of spend, rising to about 0.7% once study design is accounted for. Treat a number well outside that band as a prompt to check the method.
When one team's number beats another's, look at how each model was built before concluding the markets differ.Differences in method, math, and how often data are measured explain much of the gap here.
Apply this to established, frequently bought U.S. packaged goods.The data are 128 estimates from 22 studies through 1980, measuring short-run sales elasticity; test before extending it to new products, durables, services, non-U.S. markets, or digital media.

Evidence

The most-cited average is small: a 1% increase in ad spend moves sales about 0.2%, though adjusting for study design raises the typical estimate to about 0.7%.
Carryover is real: roughly half of one period's advertising effect carries into the next, so spending keeps working after it stops.
Leaving carryover out of a model inflates advertising's measured effect in any single period: the lingering effect of past spending gets counted as this period's response, adding about 0.34 to the short-run figure.
Skipping other sales drivers, like price or seasonality, also inflates the advertising number.
How the model handles the math, and how often the data are measured, can move the number as much as real market differences.

Key takeaway

Advertising's short-run effect on sales is small.

Source

Assmus, G., Farley, J. U., & Lehmann, D. R. (1984). How advertising affects sales: Meta-analysis of econometric results. Journal of Marketing Research, 21(1), 65–74. https://doi.org/10.1177/002224378402100107

Read the paper ↗

Evidence strength: Moderate. Based on 128 models from 22 published studies of predominantly established, frequently bought U.S. consumer products through 1980; generalizes most confidently to that setting, and less confidently to new products, durables, services, non-U.S. markets, digital media, or non-sales outcomes.