Anchoring
[Last Updated January 21st 2024]
The anchoring effect occurs when a reference point (an anchor) influences our subsequent judgement and decision making. An experiment by Tversky and Kahneman (1974) provided a clear example of this by asking individuals to estimate the number of African countries in the United Nations, after showing them a random number between 0 and 100. Those who saw the number 25 had a median estimate of 10, while those who saw the number 65 had a median estimate of 45. Tversky and Kahneman (1974) argued that this is a heuristic (a mental shortcut) where we use the first piece of information we see as an anchor or reference point, and then adjust away from it. Tversky and Kahneman (1974) also performed another experiment by asking high school students to estimate either the product of 1 x 2 x 3 x 4 x 5 x 6 x 7 x 8, or the product of 8 x 7 x 6 x 5 x 4 x 3 x 2 x 1, but only giving them five seconds to answer (not enough time to go through each number). Both answers are identical (40,320), but those estimating the first (which started with the number 1) had a median estimate of 512, while those estimating the second (which started with the number 8) had a median estimate of 2250. Tversky and Kahneman (1974) theorized that due to a lack of time, they only performed the first few steps and then used that as an anchor for their estimate. Ariely et al. (2003) performed a similar experiment using the last two digits of individual’s social security numbers to influence how much they were willing to pay for various consumer goods like wine (See “Social Security Numbers and Prices” in our Research Examples section on this page), demonstrating how important this theory is in user experience design and marketing strategy. A number of recent meta-analyses that have examined decades of studies in anchoring have found consistent moderate to large effects (Li et al., 2021; Röseler & Schütz, 2022; Schley & Weingarten, 2023), suggesting that this is a robust theory that ought to be considered in marketing strategy and user experience design as it will likely impact customer decision making. However, recent meta-analyses by Kakinohana et al. (2022) suggest that the anchoring effect may vary across cultures. Thus, this should kept in mind when considering the anchoring effect outside of North America and Western Europe.
Tversky and Kahneman (1974) believed that our problematic predictions occur due to an adjustment away from an anchor. However, Mussweiler & Strack (1999) put forth an alternative explanation called the selective accessibility model, where instead of shifting away from an anchor, we initially judge whether an anchor is the right answer to an estimate. If we don’t believe it is, then we make another guess. These additional guesses are primed (and thus influenced) by evidence considered in respect to the original anchor. Imagine if you were asked to estimate if there are more or less than 25 African countries in the United Nations (UN). The number 25 might lead you to selectively focus on information in your memory related or similar to that number. Mussweiler & Strack (1999) describe this as a form of semantic priming. For example, maybe you remember a news article about how only a small handful of African nations voted for recent anti-war legislation. This news article is now more salient and cognitively more accessible. So, if you decide “there is no way there are 25 African nations in the UN” you think of this news article, and remember that it talked about 10 or so nations. And thus, you estimate that there are 10 African nations in the UN. In this situation, you may have also learned in class how most countries are part of the UN, and might know that there are at least 40 or more countries in Africa. However, due to being primed by the number 25, these thoughts wouldn’t become salient in memory. On the other hand, if you were asked to estimate if there are more or less than 45 African countries in the UN, you may have remembered this information instead, and the news article may not have been salient. Thus instead of guessing 10, you might have guessed 40. The key here is that the anchor primes individuals to consider anchor-congruent knowledge. Mussweiler and Strack (1999) provide evidence for this selective accessibility model that you can read if interested. Epley and Gilovich (2001) however suggest that the evidence for the selective accessibility model may rely on anchors being provided externally, which is not always reflective of reality. Rather, they suggest that in many situations we generate our own anchors. For example, if you were asked to estimate how many African nations were in the UN without being provided an external anchor, you might start by thinking “well there are at least 40 countries in Africa” and then adjust from there. Epley and Gilovich (2001) ran three studies to test this, and found that self-generated anchors often result in adjustment (e.g. 73.9% of the time in experiment one), while experimenter-generated anchors led to various other forms of reasoning (e.g. adjustment from an experimenter-generated anchor only being used 13% of the time in experiment one). Thus, one takeaway from these theories and studies is the importance of considering how anchors are communicated and perceived.
To begin with, consider where anchoring would be appropriate. Do you want to influence customers who are estimating a value related to your brand (like popularity, sales, customers, social media followers)? Or are you looking to impact pricing? You might also want to influence how values are perceived in general (e.g. influence whether a number feels like a lot or a little). A good rule of thumb is to consider how anchoring might affect decision making whenever numbers might be involved. Once you have a general idea of how anchoring may come into play, you want to consider whether you will be providing the anchor, or you will be trying to influence the customer to come up with an anchor on their own. If you are providing the anchor, you may want to write out how a customer might interpret the provided anchor in respect to the selective accessibility model. On the other hand, if the customer is generating the anchor on their own (even if it’s based off non-numerical primes you provide), you may want to think about how they might adjust away from the anchor. You also want to consider if the anchors being provided or perceived are realistic or unrealistic. If they are unrealistic, you may want to consider changing them to something that the customer may perceive as being an accurate value (See “Social Security Numbers and Prices” in our Research Examples section on this page for the reasons why). Further, you want to consider any other possible heuristic impacts that may arise in conjunction with anchoring. For example, if you are using multiple tiers of pricing on a website, you may want to consider if you are engaging the decoy effect. Or if your anchors are being used to influence how much individuals are willing to trade/sell an item for, you might want to consider the impacts of the endowment effect and loss aversion. You also need to consider demographic differences. Culture may impact how individuals interact with anchors. And socioeconomic status may influence perceptions and interactions with anchors used in pricing. As anchoring studies are extremely varied, there are many more considerations to keep in mind. Doing research on anchoring in respect to your specific goals or industry may be helpful, given the hundreds of different studies available for review.
SAAS businesses often operate on pricing plans, and if you’ve ever downloaded a WordPress theme, you know that these tend to come in threes. Determining what three service plans to use can take some experimentation. However, there are a few tricks you can consider that take advantage of anchoring. The first trick is to use the first price to set expectations (assuming the first choice isn’t a free plan). Alternatively, you can use the middle price as an anchor by making its container on a website stand out (make it larger, bolder, a different colour, or indicate that it’s the most popular choice). These anchor prices can inform visitors to your website about the value of your service, especially if it is unique. Knowing this, you don’t want to set a large gap between the anchor and other tiers offered, as it may artificially devalue those tiers. For example, if tier one is $10, and tier two is $100, customers may feel like tier two doesn’t provide the same “bang for the buck” as tier one. You can also use a middle option as an anchor that drives sales of the highest tier of service. For example, you might have a $70 tier that has 3 features (but the least needed features available), a middle tier for $100 that has 5 features (with the most in-demand feature), and the top tier that has 10 features for $110 (with the five most in-demand features). In this situation, you would want to emphasize the middle tier. Customers would mostly ignore the introductory tier because it doesn’t have the feature they need. As the middle tier is emphasized, they would use it as an anchor. And they would compare it to the top tier, which has considerably more features for only $10 more, making it seem like a great deal. You could also create a similar strategy using the decoy effect. For example, tier one might have 3 features for $50 a month, tier two might have 6 different features from tier one for $80 a month, and tier three might have 9 features “on sale” for $80 a month. Here, the middle feature (which you would once again emphasize) acts as an anchor, and as a decoy (as it’s an inferior option), which makes the third tier look like a great choice, even if the user only really needs the first tier.
Wansink et al. (1998) ran a study across 86 grocery stores and found that multi-unit pricing led to a 32% increase in sales over single-unit pricing. Thus, when running a sale, you can use the quantity of what you are selling as an anchor. For example, instead of saying “Soda on sale for $0.25 per bottle” you can sell more if you say “4 bottles of Soda on Sale for $1.” Similarly, if running an e-commerce store, you can sell products using multi-unit pricing rather than selling single units and asking customers to select a quantity. You will notice this practice often occurs on Amazon. Keep in mind that this works better for some products than others, and should be tested. For example, low-calorie chocolate snack bars may not sell well if multi-unit pricing is used, as they tend to be targeted towards health-conscious customers as impulse purchases. On the other hand, any product that people might consider buying multiples of would likely benefit from multi-unit pricing.
Wansink et al. (1998) ran a study on discounted cans of soup at three grocery stores, and found that quantity limits can act as an anchor that individuals adjust down from, which in turn increases sales. For example, they found that when soup was discounted by 10 cents, no limit led to an average of 3.3 cans purchased, a limit of 4 led to an average of 3.5 cans purchased, and a limit of 12 led to an average of 7 cans purchased. Thus, when running a sale, you should consider setting a high limit to act as an anchor. For example, if selling soda for 20% off, you might set a limit of 10 bottles. In this situation, the adjustment model suggests customers would adjust down from 10, rather than up from 0, leading to more sales than if you did not use a limit. On the other hand, if the selective accessibility model plays a role in this situation, the limit would act as an anchor to bias the customer’s decision making process. For example, they might ask themselves “do I need 10 bottles of soda” and then selectively elicit information from their memory that supports their need for many bottles of soda (e.g. “well my kids drink a lot of it, and it disappears fast”). If instead there was no limit or anchor, they would not be primed to think of why they might need more than one or two bottles.
When launching a brand new product/service that individuals are unable to compare to other alternatives, you have an opportunity to use anchoring effects to set your price. If you can get individuals to estimate that the value of your product is higher than the actual cost (e.g. through advertisements), their estimate (anchor) will make them more likely to purchase your product (as it will seem like a better deal). Essentially, you can use your branding to create a perceived value (an anchor), and then price your product lower than said anchor (e.g. a “fair/good price”).
Many ecommerce sites show an initial value crossed out, with a sale value beneath it. That initial crossed out value acts as an anchor price, making the sale price feel like a better deal. However, you need to be careful to make sure that your trick isn’t obvious. If you price something at $800 and have it for sale at $50, most customers will assume you’re scamming them, or selling something illegal/stolen/bootlegged.
Ariely et al. (2003) had participants from an MBA market research course look at products in six different categories, such as computer accessories and wine bottles. They were then asked if they would purchase the products for an equal dollar value to the last two digits of their social security number. Following this, they were asked what the maximum value they would pay for the item was. They were also told there was a random chance that their answers would lead to an actual transaction (where they had to purchase a product at either of the two prices). Ariely et al. (2003) found that those above the median social security number were willing to pay 57% to 107% more for products than those under it. Further, when they split the social security numbers into quintiles, those in the top quintile (e.g. highest numbers) were willing to pay up to three times more than those in the lowest. This demonstrates how a reference point (anchor) can drastically influence an individual’s willingness to pay for a product. However, when presented with different products in a category, nearly everyone ranked their values similarly. For example, participants were nearly all willing to pay more for a good bottle of wine than a cheap bottle of wine. Thus, when considering pricing strategy, it’s important to consider that customers will still differentiate between options. And thus, anchors you use will likely influence all prices. For example, if you offer three tiers of monthly membership for software as a service, the first price people see will likely influence their perceptions of all other prices. Usually the lowest tier is listed first. And thus, the lowest tier you choose may anchor expected values. And this can potentially create issues if there is a large gap in cost between the lowest option and the higher tiers. This is why you will see some websites make the middle option more salient through design (e.g. making the middle option stand out more). It is important to note however that other studies have been unable to replicate this when random numbers were used. Fudenberg et al., (2012) performed the experiment on mainly undergraduate students using excel to randomly generate a number, but asked participants about their willingness to sell rather than their willingness to pay. They did not find that the random numbers had a strong anchoring effect. But, by telling participants they could potentially keep the good, they were likely engaging the endowment effect and loss aversion, two heuristics that may have overrode the anchoring effect due to negativity bias. Fudenberg et al., (2012) then ran a second experiment that looked at willingness to pay, but this would also be problematic as an undergraduate population would not likely have the budget to afford to actually purchase a good, and thus would be more likely to underestimate any price, potentially overriding any anchoring effects. MBA students on the other hand are more likely to have the wealth to afford low-cost items used in the Ariely et al. (2003) experiment. This may demonstrate that the anchoring effect is more likely to occur when making an automatic choice/estimate (e.g. in the case of MBA students), than when overriding automatic processes to cognitively consider what one can afford (e.g. in the case of undergraduate students). Which in turn, would be in line with expectations regarding heuristics. And may also suggest that anchoring works best in situations where the pricing is affordable and in line with expectations, or in situations where the product/service is completely novel and pricing can’t be compared to alternatives (e.g. when Midjourney AI first launched). Ionnidis et al. (2020) ran a similar experiment, where they had participants generate their number by rolling a dice, and were also unable to replicate Ariely et al.’s (2003) results. However, they used willingness to sell rather than willingness to pay leading to the same issues as Fudenberg et al.’s (2012) first experiment. Further, they followed this up with a meta-analysis that considered studies using willingness to pay as they recognized that the effect had previously been stronger for it. They found no difference between studies that used willingness to sell (technically willingness to accept, but for simplicity we are calling it willingness to sell here) and willingness to pay, but did find that anchoring tended to play a role in studies where the number provided wasn’t perceived as purely uninformative. Ionnidis et al. (2020) suggest that people are likely being tricked into thinking uninformative numbers (e.g. a random number) are informative (e.g. somehow meaningful in respect to the estimate they are making), and thus the anchoring bias is really a perception bias. If this is the case, it suggests that when taking advantage of anchoring, you need to make sure that the numbers you use are somewhat relevant to the value expectations you want consumers to estimate. In the previous example we provided of a three-tiered monthly membership, the values you choose for pricing each tier would all be relevant to each other, and thus would likely be influenced by anchoring. Ionnidis et al. (2020) also found that when numbers were perceived as questionable or uninformative, anchoring was strongest for novel goods where participants were unable to compare pricing to existing/familiar goods.
Ariely, D., Lowenstein, G., & Prelec, D. (2003). “Coherent arbitrariness”: Stable demand curves without stable preferences.
The Quarterly Journal of Economics, 118(1), 73-106.
https://doi.org/10.1162/00335530360535153
Epley, N., & Gilovich, T. (2001). Putting adjustment back in the anchoring and adjustment heuristic: Differential processing of self-generated and experimenter-provided anchors.
Psychological Science, 12(5), 391-396.
https://doi.org/10.1111/1467-9280.00372
Fudenberg, D., Levine, D. K., & Maniadis, Z. (2012). On the robustness of anchoring effects in WTP and WTA experiments.
American Economics Journal: Microeconomics, 4(2), 131-145.
https://doi.org/10.1257/mic.4.2.131
Ioannidis, K., Offerman, T., & Soof, R. (2020). On the effect of anchoring on valuations when the anchor is transparently uninformative.
Journal of the Economic Science Association, 6, 77-94.
https://doi.org/10.1007/s40881-020-00094-1
Kakinohana, R. K., Pilati, R., & Klein, R. A. (2022). Does anchoring vary across cultures? Expanding the many labs analysis.
European Journal of Social Psychology, 53(3), 585-594.
https://doi.org/10.1002/ejsp.2924
Li, L., Maniadis, Z., & Sedikides, C. (2021). Anchoring in economics: A meta-analysis of studies on willingness-to-pay and willingness-to-accept.
Journal of Behavioral and Experimental Economics, 90. 101629.
https://doi.org/10.1016/j.socec.2020.101629
Mussweiler, T., & Strack, F. (1999). Hypothesis-consistent testing and semantic priming in the anchoring paradigm: A selective accessibility model.
Journal of Experimental Social Psychology, 35(2), 136-164.
https://doi.org/10.1006/jesp.1998.1364
Röseler, L., & Schütz, A. (2022). Hanging the anchor off a new ship: A meta-analysis of anchoring effects.
PsyArXiv.
https://doi.org/10.31234/osf.io/wf2tn (NOTE: This is a preprint and not peer reviewed)
Schley, D. R., & Weingarten, E. (2023). 50 years of anchoring: A meta-analysis and meta-study of anchoring effects.
SSRN.
https://dx.doi.org/10.2139/ssrn.4605057 (NOTE: this is not technically peer reviewed in the sense that a journal required peer review for publication, but it does come from a legitimate source)
Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases.
Science, 185(4157), 1124-1131.
https://doi.org/10.1126/science.185.4157.1124
Wansink, B., Kent, R. J., & Hoch, S. J. (1998). An anchoring and adjustment model of purchase quantity decisions.
Journal of Marketing Research, 35(1), 71-81.
https://doi.org/10.2307/3151931
Discover More Knowledge
Contact Us