Survey research can be subject to many sorts of biases. One of the most interesting ones in brand research is acquiescence bias — the tendency of some respondents to lean towards providing a positive response to questions, and to thus, for example, falsely say that they know a brand or would consider buying it (or have already done so) because they feel that it’s socially desirable to do so.

The impact of acquiescence bias on data quality can be quite profound. It is not uncommon for survey results to show that 5% of the respondents claim to have purchased a product, whereas actual sales figures indicate that the true share should be closer to 0.1%. This kind of overclaim can be especially problematic for questions branched from another, as is often the case with funnel KPIs from awareness in brand surveys. If people who falsely claim to know a brand are asked more in-depth follow-up questions about it, they are unable to answer them, which introduces bias to the results.

This impact is not necessarily consistent across brands either — it can be greatly exaggerated for lesser-known brands than those with very high awareness. Small brands are most impacted by acquiescence bias, given that even small errors in reported awareness can translate into quite drastic percentage leaps and even bigger representations of people in real population terms.

To better understand the impact of acquiescence bias on our surveys, we designed a test to capture overclaim across various countries. In this article, we’ll explore the findings and their impact further.

How can we test the strength of acquiescence bias in our data?

We tested an effective way to capture over-claim behaviour using a fake brand question within our usual brand surveys. This mimics the exact design and placement of a brand awareness question and so cannot easily be detected as a test question. Respondents who claim to know the fake brand we present to them are considered to show over-claim behaviour, a clear indication of acquiescence bias.

The design of the logo used to depict the fake brand is an important element in the effectiveness of this capture.  Ideally, we needed to create a logo and name that is coherent with the rest of the brands in the survey, but not so subtle as to easily be mistaken as a real brand. The sweet spot is creating a fake brand that is unique and could be real but isn’t! We developed dozens of these fake brand questions that we randomly attributed to respondents answering our surveys — this approach ensured that we minimise the effect of any one fake brand being more or less effective than another, or from any potential bias from one fake brand accidentally resembling a real brand.

To be even more sure we don’t capture only accidental awareness of these fake brands, we allowed respondents the option to provide a non-answer to the fake brand question. Including a “not sure” answer option enabled respondents who are uncertain to provide an honest answer to the question, as opposed to forcing them into providing a binary response. For the purpose of this test, we are only evaluating those respondents who have answered “yes” to knowing our fake brand, as this is the most conclusive evidence of bias.

We tested our fake brand questions over 2 months (December 2024 to January 2025), asking over 860,000 respondents across different countries if they were aware of the fake brand randomly attributed to their survey.

How does acquiescence bias differ by country, demographic or socioeconomic context?

Our testing spanned 20 markets and showed a lot of variance in acquiescence bias rates between them. Japan, Norway, Denmark, France and Sweden all had the lowest rates of bias, with less than 3% of people claiming to know the fake brand. Middling countries like Germany, Belgium, Poland, Brazil, Portugal and Switzerland saw rates of over-claim between 5–8%. Those countries with the highest rates of over-claim on the fake brand question by far were Israel (13%), India (14%) and China (15%).

Beyond cultural or country context, we also see differences in demographics like age or gender, as well as socioeconomic background.

Firstly, if we look at age groups, we can see that younger demographics tend to show more acquiescence bias than older respondents, with those aged 18–25 being the highest over-claimers. If we cross-examine this by country, the most notable spike in acquiescence bias by younger demographics is seen in the US, Germany and also Israel. Some countries defy this pattern — in Portugal and China, it is the older demographics who present the most over-claim behaviour. France, the Netherlands and Switzerland all show a higher acquiescence trend for those aged 26–35, too.

We also see males tend to over-claim on the fake brand question more so than females, by 65%. This is conclusive across all 20 of the countries surveyed, showing a clear gender differentiation, however, this is most differentiated in China, Israel and Portugal too. Some countries showed no difference across gender, as is the case in Japan and the Netherlands, while both Norway and Sweden showed a higher acquiescence bias rate for women than men.

Income levels also show a consistent pattern in regards to acquiescence bias rates, in that as income increases, so does the rate at which respondents say “yes” to fake brand awareness. This is most notable in China and India, as well as in Germany, Belgium and Poland. In Norway and Israel, the pattern is reversed, however, with lower-income respondents more likely to exhibit over-claim behaviour.

Education levels are less conclusive in terms of a pattern of over-claim behaviour overall, though there are some differences across countries: in India, China and Portugal, respondents with a lower education tend to show higher acquiescence bias than others. Great Britain, the Netherlands, Belgium and Brazil all show no difference between education levels. And Germany and Switzerland show higher over-claims from respondents from a higher education background.

How does correcting for acquiescence bias impact the resulting data?

Given that we saw such vast differences in acquiescence bias across countries, we then wanted to explore the impact this has on the resulting brand data. We tested the awareness results of sports brands on 1,000 people in India, a country where we measured one of the highest rates of acquiescence bias:

For the two very well-known brands, Nike and Puma, the impact was fairly minimal, with only small declines in awareness levels once the acquiescence bias is removed from the results. However, as the brand awareness decreases, the distance between the pre-cleaned and post-cleaned results drastically increases. For Skechers, we see a drop of 7% in reported awareness levels after the removal of respondents with an acquiescence bias, and for New Balance and Saucony, it was as high as 10% and 11%, respectively. In the case of Saucony, almost two-thirds of the reported awareness level seems to have been due to overclaim.

The impact on the data when translated to real population numbers is staggering, especially for a market like India, where a few percentage points of difference translates into tens of millions of people who are (or aren’t) aware of a brand. Even for the minimally-impacted brands like Puma, the drop in awareness level once the acquiescence bias is removed represents 9.4 million people. For brands like Saucony, the effect could be as big as 100 million people falsely claiming to know it. 

The impact on the data when translated to real population numbers is staggering, especially for a market like India, where a few percentage points of difference translates into tens of millions of people who are (or aren’t) aware of a brand. Even for the minimally impacted brands like Puma, the drop in awareness level once the acquiescence bias is removed represents 9.4 million people. For brands like Saucony, the effect could be as big as 100 million people falsely claiming to know it.

Final Thoughts

1. Have better comparability across countries

Acquiescence bias is an unavoidable consequence of asking people survey questions, but removing it is key to obtaining high-quality insights that are comparable across countries. We decided to implement the fake brand question into every survey we run, enabling us to:

2. Ensure that key metrics are not inflated, especially for small brands

For most of our clients, it’s not just important to have highly accurate trendlines, but also to be able to compare their KPIs across countries in a consistent way. Removing acquiescence bias enables us to create meaningful benchmarks, e.g. Japan and China, even though the latter has overclaim rates that are more than 5 times higher than the former.

3. Increase the accuracy of funnel metrics

Any misrepresentation of key metrics, like brand awareness, is frustrating for tracking changes over time with any accuracy,  and this is especially exaggerated when it comes to brands with lower awareness levels, where any misstep in the data is far more pronounced in the end result. Eliminating over-claim ensures that any depicted trends in key metrics like awareness are real.

Beginning with more accurate awareness will ensure that any following brand metrics are measured on the right people who are capable of answering questions about the brand. This eliminates the risk of sample-based skews that offset the calculations of funnel KPIs, like brand consideration or perception, or strange answer behaviour by those who don’t know the brand enough to answer follow-up questions about it.

‍