Guides · Methods and evidence

Effect sizes and meta-analysis in marketing research

Marketing reports effects mainly as elasticities, correlations and standardized mean differences, and each answers a different question. Meta-analysis converts study results to a common metric, weights them by precision, and uses moderators to explain why they differ, which is often as informative as the average itself. Before quoting a pooled figure, check its unit and horizon, the method moderators, the spread of effects and whether publication bias was tested.

Three metrics and what each one tells you

Market response studies usually report elasticities: the percentage change in sales or market share for a 1% change in a marketing variable such as advertising or price. Because elasticities are unit-free, estimates from different brands, categories and decades can be compared and averaged, and short-term and long-term elasticities are reported separately because they answer different questions.^[1][2]

Survey-based research is usually pooled as correlations. In a meta-analysis of market orientation and business performance, the Pearson correlation was the metric most studies reported, so it became the common unit. Large replication projects also convert test statistics into correlations where possible, because correlations are bounded, familiar and, after the Fisher transformation, have a standard error that depends only on sample size.^[3][4]

Experiments are usually summarized as Cohen's d, the difference between group means divided by the pooled standard deviation, which makes results comparable across studies with different response scales. A meta-analysis of cause-related marketing experiments used this metric and noted the conventional labels of .2 for a small and .5 for a medium effect.^[5]

Benchmarks from marketing meta-analyses

Published meta-analyses give reference values against which a new estimate can be judged. Each is the average of the studies it pooled, estimated with the methods available at the time.^[1][6]

• Advertising: the first meta-analysis pooled 128 econometric models and put the mean short-term elasticity at about .22; a 2011 update of 751 short-term and 402 long-term brand elasticities found means of .12 and .24.^[7][1]
• Price: across 1,851 price elasticities from 81 studies, the mean was minus 2.62, far larger in absolute size than the advertising averages.^[2][1]
• Personal selling in pharmaceuticals: 373 detailing elasticities from 48 papers averaged .21, falling to .18 after correction for method-related biases.^[8]
• Firm value: 488 stock-market elasticities from 83 studies averaged .04 for advertising spending and .54 for marketing assets, with customer-related assets at .72 and brand-related assets at .33.^[6]
• Online word of mouth: 1,532 effect sizes from 96 studies gave an average correlation with sales of .091, with volume mattering more than valence.^[9]
• Market orientation: across 56 studies of 14,586 firms in 28 countries, correlations with performance were about .25 for business-level and .26 for market-specific measures.^[3]

From an effect size to a decision

An elasticity becomes useful once it is combined with economics. The pharmaceutical detailing meta-analysis took its bias-corrected elasticity, combined it with plausible product margins, and derived how detailing spending should shift as products age and how it should differ between Europe and the United States.^[8]

Averages can also hide useful contrasts. In the firm-value meta-analysis, advertising elasticities were lower in more concentrated industries and marketing-asset elasticities were higher during recessions. In the ad repetition meta-analysis, brand attitude peaked at around ten exposures in experimental settings while recall kept rising through at least the eighth exposure, so the answer to how many exposures are enough depends on which outcome matters.^[6][10]

How a meta-analysis is built

The core steps are the same across marketing meta-analyses, although the details differ by field.^[3][11]

• Define the construct and inclusion rules. The market orientation meta-analysis admitted only studies that used items from the two established scales and excluded non-profit organizations, so that every effect referred to the same relationship.^[3]
• Search systematically and code each result into a common metric, which requires that studies report sample sizes and correlations or statistics that can be converted.^[3][4]
• Correct for statistical artifacts where appropriate. The transaction cost meta-analysis weighted correlations by sample size and corrected them for artifacts before pooling.^[11]
• Choose the pooling model. A fixed-effects model assumes one population effect plus sampling error; a random-effects model assumes true effects differ across studies and adds that variation to each study's weight.^[12][13]
• Test for heterogeneity and explain it with moderators or meta-regression, which yields expected effects for particular settings rather than one number.^[11][13]
• Test for publication bias and report the adjusted estimate alongside the original one.^[11][13]

Why estimates differ: moderators and study methods

The earliest advertising meta-analysis treated published studies as imperfect replications of one unplanned experiment and used analysis of variance to find what explained their differences. Short-term elasticities varied with model specification, estimation method, measurement, product type and study setting, so much of the spread in a literature comes from how effects were estimated, not only from how markets differ.^[7]

Later work kept confirming this. Advertising elasticities were higher for durables, for products early in their life cycle, with yearly rather than quarterly data and when advertising was measured in rating points rather than money, and leaving lagged sales or distribution out of the model inflated them. For price, allowing for price endogeneity made elasticities larger, while allowing for consumer heterogeneity made no significant difference.^[1][2]

Measurement matters in survey literatures too. Studies that assessed performance subjectively found stronger links with market orientation than studies using objective data, with corrected correlations of about .28 against .22, and the scale used to measure market orientation also changed the effect. The same meta-analysis found weaker effects in countries culturally distant from the United States, a contingency no single-country study could have shown.^[3]

Publication bias and the pooled average

If small or nonsignificant estimates are less likely to be published, the pooled mean is inflated; bias corrections cut the average advertising elasticity by more than a factor of five in a 2025 re-analysis and removed the evidence for an overall effect in a large nudge meta-analysis.^[13][14]

Correction does not always change the conclusion. In the transaction cost meta-analysis, trim-and-fill analysis suggested no missing studies for 11 of 14 relationships, and the adjusted estimates left every conclusion intact. The aim is to test for bias and report what the test shows, not to assume the answer in either direction.^[11]

Checks before you cite or report a pooled effect

These questions apply whether you are quoting a meta-analytic average or reporting a new estimate that others will pool later.^[7][15]

• Check the unit and horizon. A short-term elasticity, a long-term elasticity and a correlation cannot be compared directly.^[1]
• Look for method moderators before trusting the grand mean, and use the conditional estimate that matches your setting.^[7][13]
• Look at the distribution of effects, not only the mean. In one re-analysis, nearly two-thirds of individual effects fell below the pooled mean.^[16]
• Compare with randomized evidence where it exists. In split-cable television tests, heavier advertising weight produced a significant sales gain in 33% of tests for established brands and 55% for new products, even at a lenient confidence level.^[17]
• Report what future meta-analysts need: sample sizes, standard errors or zero-order correlations, and how each construct was measured. In the market orientation meta-analysis, 56 of about 175 candidate empirical studies reported enough to be included.^[3][15]

The evidence behind this

• Empirical generalizations in marketing: what the research shows
• How advertising works: mechanisms, elasticities and returns
• Marketing and firm value: what the research shows
• Research methods pitfalls: ratio variables and censored data

References

1. Raj Sethuraman, Gerard J. Tellis, Richard A. Briesch (2011). How Well Does Advertising Work? Generalizations from Meta-Analysis of Brand Advertising Elasticities. Journal of Marketing Research.
2. Tammo H. A. Bijmolt, Harald J. van Heerde, Rik G. M. Pieters (2005). New Empirical Generalizations on the Determinants of Price Elasticity. Journal of Marketing Research.
3. Paul D. Ellis (2006). Market Orientation and Performance: A Meta‐Analysis and Cross‐National Comparisons*. Journal of Management Studies.
4. Open Science Collaboration (2015). Estimating the reproducibility of psychological science. Science.
5. Christina Schamp, Mark Heitmann, Tammo H.A. Bijmolt, Robin Katzenstein (2022). The Effectiveness of Cause-Related Marketing: A Meta-Analysis on Consumer Responses. Journal of Marketing Research.
6. Alexander Edeling, Marc Fischer (2015). Marketing's Impact on Firm Value: Generalizations from a Meta-Analysis. Journal of Marketing Research.
7. Gerald Assmus, John U. Farley, Donald R. Lehmann (1984). How Advertising Affects Sales: Meta-Analysis of Econometric Results. Journal of Marketing Research.
8. Shrihari Sridhar, Murali K. Mantrala, Sönke Albers (2013). Pharmaceutical Detailing Elasticities: A Meta-Analysis. International Series in Quantitative Marketing.
9. Ana Babić Rosario, Francesca Sotgiu, Kristine De Valck, Tammo H.A. Bijmolt (2015). The Effect of Electronic Word of Mouth on Sales: A Meta-Analytic Review of Platform, Product, and Metric Factors. Journal of Marketing Research.
10. Susanne Schmidt, Martin Eisend (2015). Advertising Repetition: A Meta-Analysis on Effective Frequency in Advertising. Journal of Advertising.
11. Inge Geyskens, Jan-Benedict E. M. Steenkamp, Nirmalya Kumar (2006). Make, Buy, or Ally: A Transaction Cost Theory Meta-Analysis. Academy of Management Journal.
12. Cynthia Rodriguez Cano, François A. Carrillat, Fernando Jaramillo (2004). A meta-analysis of the relationship between market orientation and business performance: evidence from five continents. International Journal of Research in Marketing.
13. Joseph Korkames, T. D. Stanley, Stefan Stremersch (2025). Meta-analysis of advertising effectiveness: New insights from improved bias corrections. International Journal of Research in Marketing.
14. Maximilian Maier, Frantis̆ek Bartos̆, T. D. Stanley, David R. Shanks, Adam J. L. Harris, Eric‐Jan Wagenmakers (2022). No evidence for nudging after adjusting for publication bias. Proceedings of the National Academy of Sciences.
15. Peter N. Golder, Marnik G. Dekimpe, Jake T. An, Harald J. van Heerde, Darren S. U. Kim, Joseph W. Alba (2022). Learning from Data: An Empirics-First Approach to Relevant Knowledge Generation. Journal of Marketing.
16. Jonathan Z. Bakdash, Laura R. Marusich (2022). Left-truncated effects and overestimated meta-analytic means. Proceedings of the National Academy of Sciences.
17. Leonard M. Lodish, Magid Abraham, Stuart Kalmenson, Jeanne Livelsberger, Beth Lubetkin, Bruce Richardson, Mary Ellen Stevens (1995). How T.V. Advertising Works: A Meta-Analysis of 389 Real World Split Cable T.V. Advertising Experiments. Journal of Marketing Research.

Bibliographic details only. Follow the DOI to read each study at its publisher.