Anchor effects in appraisals: Does educational background matter?

In both everyday and professional contexts, individuals often base their judgments on an initial value or perception, subsequently adjusting this reference point to arrive at a final decision. The initial value, or starting point, may be derived from a problem formulation, an initial calculation, previous experience or a first impression. In all cases, adjustments based on these initial perceptions are rarely sufficient. This means that different starting points or initial values produce different estimates and thus also different outcomes or judgments. These are therefore biased because they are affected by the initial values. We call this phenomenon the anchoring effect (Slovic and Lichtenstein, 1971). One profession that is subject to this phenomenon is the property valuation.

In the everyday work of the property valuer, factors such as market trends, economic development, financing, and the characteristics and location of the property are essential for assessing market value. In addition, the appraiser must also take into account psychological factors affecting both the market and him. Among these psychological factors, anchoring can be a factor that complicates the assessment of market value. Moreover, to add to the complexity, this information does not have to be relevant or even related to the subject matter, which the classic experiment of Tversky and Kahneman (1974) proved. This contrasts with the assumption that professionals in the real estate valuation industry act independently based on full and relevant information, have rational and consistent preferences and are able to identify factors influencing the value. However, in property valuation several studies demonstrate that judgments can be irrational (Cheloti and Mooya, 2024; Kokot, 2023). At the same time, property valuation constitutes an important cornerstone in our society. On the individual level, we are dependent on estimates of property values when buying, selling and financing our homes. On the company level, we use valuations not only in the transaction process and for financing, but also for reporting. For many companies the estimated property values constitute a large part of the company book value (Lind and Nordlund, 2021). Waszczuk (2024) demonstrates that anchoring effects are not limited to professional valuers but also influence households' price expectations. Her study on the Warsaw housing market shows that even irrelevant external information can significantly shift valuation judgments, suggesting an autoregressive way of thinking among consumers. A notion that is confirmed in a recent experimental study by Węgrzyn and Kuta (2024). They found that even uninformative, randomly assigned numbers significantly influenced participants' estimations of asking and transaction prices. Their findings suggest that anchoring effects can be triggered without any meaningful reference point, reinforcing the idea that valuation judgments are highly susceptible to cognitive framing – even among students with real estate training. Taken together, this line of research, Lind and Nordlund (2021), Waszczuk (2024), Węgrzyn and Kuta (2024), supports the notion that anchoring is a pervasive cognitive bias in real estate decision-making that bear real consequences for society at large, as the real estate market has become a fundamental market linking individuals, households, credit market, government and institutions together (Palm and Andersson, 2021).

The rest of the paper is structured as follows. First, a theoretical framework is outlined to frame the research. Second, the purpose and hypotheses are developed from the theoretical framework and the concept of anchoring in property valuation. Third, the design of the study is outlined with the details of the experiment conducted. Subsequently, the results are presented together with their analysis. Lastly, we discuss the results and relate them to previous studies and theory. Implications for further study and practice are included in this section.

The issue of a valuer's misjudgment is often attributed to the adoption of cognitive heuristics (Diaz, 1997; Gallimore, 1996). Heuristics are cognitive shortcuts used in complex problems-solving or decision-making (Simon, 1955). As complexity and the amount of detailed information increase, people prefer heuristics to eliminate alternatives, often with limited information search and evaluation (Gallimore et al., 2000). This approach reduces the time and effort required for valuers, lowering transaction costs associated with searching for and processing information. Similar findings have been reported in developing countries, where valuers frequently rely on heuristics to cope with limited market information. Cheloti and Mooya (2024) show that while anchoring and adjustment heuristics can improve valuation accuracy in information-scarce environments, their inappropriate use contributes to valuation inaccuracies. Their study highlights the dual nature of heuristics – both as a cognitive shortcut and a potential source of bias.

There are different kinds of heuristics. The Availability Heuristic, as denoted by Tversky and Kahneman (1973) indicates that individuals evaluate the frequency or probability of events based on their availability. Quan and Quigley (1991) show that valuers rely on memory, past experiences, beliefs, principles, preferences, perceptions, intuition, sentiments, interpretation, and human capital when valuing a property. Information derived from internal reflection and self-awareness tends to be more immediately accessible and easily retrieved compared to macroeconomic, market, and property-specific data and characteristics. Valuers prefer the most recent or vividly recalled information, leading to recency bias and anchoring (Baum et al., 2000; Diaz et al., 2002).

The representative heuristic is stereotyping, where valuers classify an event with others they are familiar with. MacCowan and Orr (2008) suggest that valuation decisions are biased towards markets familiar to valuers and those with good-quality data. This heuristic also applies to herding behavior, where valuers follow the majority by relying on their valuation information rather than conducting independent analysis. This behavior is accentuated in high-uncertainty decisions involving heterogeneous assets or technical knowledge in the valuation process.

Valuers often solve problems by forming a priori value estimates as a reference (Iroham et al., 2014) which we can relate to anchoring and adjustment heuristic. This process can involve personal contacts, expert opinions, price asking techniques, or relying on incomplete contract prices, sensationalist media, or previously appraised values (Diaz et al., 1999; Clayton et al., 2001). Overconfidence can result from overreacting to confirmed private information, leading to erroneous market interpretations (Gallimore et al., 2000).

Evans (1993) also noted that humans tend to seek information consistent with their current beliefs and avoid potentially falsifying evidence, which we label as confirmation bias. Valuers look for ways to confirm their valuation perceptions, often eliminating or underreacting to contrary evidence (Gallimore, 1994; Baum et al., 2000). This bias can lead to the manipulation and adjustment of existing information to fit and confirm their views.

The selection of data for valuation is not purely mathematical but a heuristic process influenced by subjective preferences, professional intuition, and gut feelings (Klein and Kahneman, 2009). These biases and heuristics, along with client influences, contribute to significant discrepancies and variations in property valuations (French and Mallinson, 2000; French and Gabrielli, 2003). Understanding these cognitive shortcuts is crucial for improving the accuracy and reliability of property valuations.

When appraising the market value, the comparable sales method is the most common valuation method (Bengtsson, 2023; Shapiro et al., 2019). However, the method has two major “problems”, similarity of comparable sales and time-lag. The problem of similarity concerns the fact that two properties cannot be located exactly at the same location and that each property is unique in at least this aspect. Furthermore, all properties will have their unique standard and features as they have been refurbished and maintained differently over time. Here, it is the individual appraiser's judgment that will define which sales will be considered comparable. The same goes for the other problem, about the time-lag. As the appraiser can only observe the transaction in retrospect and the real estate market is rather illiquid, it is up to the individual appraiser to judge how far back in time the sales should be treated as comparable. George et al. (2000) studied how decision support systems, which are common within the real estate valuation industry, can mitigate anchoring. However, they found that the anchoring effect were persistent, and the support system did not have any countering effect. With this in mind, Appraisal Institute (2020) concluded that the final value opinion is not the average of the different value indications derived. No mechanical formula should be used to select one indication over the others, rather, final reconciliation relies on the proper application of appraisal techniques and the appraiser's judgment. This calls for sound judgment in a rational process from the appraiser's side. But Newell and Simon (1972) states that the human being cannot make perfect rational decision. This leaves us with the different heuristics and a profession where we to a high degree relies on the judgment of the individual appraiser.

Bellman (2022) argues that students during their education develop cognitive thought patterns influencing their professional socialization as defined by Hofstede et al. (2010), Levy and Frethey-Bentham (2010). Bellman et al. (2016) demonstrate in their study that different property education programs influence the students thought patterns differently. One conclusion is that engineering programs with a more quantitative tradition are more model-driven. This as the education program are more standardized and norm-driven, with a focus on mathematics. Furthermore, they state that the normative education tradition is known to present information in “binary terms”, in contrast to social science-based education programs that allows more degrees of freedom in interpretations of information, leading to wider thought patterns.

Therefore, it is of great interest to test if this difference in thought patterns also is applicable to heuristics in terms of anchoring. The aim of the study is to investigate how anchoring is manifested in property educational programs in either an engineering or a social science context.

This difference in educational framing may influence susceptibility to cognitive biases such as anchoring. Palm and Andersson (2021) found that students provided with a low anchor on average appraised a lower value of the property than those provided with a high anchor, even after being informed about anchoring effects. This suggests that theoretical knowledge alone may not be sufficient to mitigate anchoring bias. Their study highlights the importance of cognitive capacity and critical thinking in valuation tasks. These findings support the hypothesis that students from social science backgrounds – who may be trained in more interpretative reasoning – are more susceptible to anchoring than those from engineering backgrounds.

In relation to these previous studies, the focus on anchor effects, and theoretical knowledge to mitigate anchoring bias, the present study develops two hypotheses:

As Bellman et al. (2016) conclude, the educational background among Swedish valuers has an influence on the cognitive thought patterns. One significant difference is that students with an engineering background are more systematically trained in model thinking. This could imply that students with an engineering background would not have the same thought patterns as students with a social science background and therefore there would be a difference in the anchoring effect between them. As Palm and Andersson (2021) found an anchoring effect among social science students, one could think that the difference in thought patterns among students with an engineering background could mitigate the anchoring effect.

The present study serves two important purposes: firstly, to contribute to the research and understanding of anchoring effects in property valuation, and secondly, to highlight the presence of an anchor effect in conjunction with educational background within the real estate industry.

The present study features an experiment conducted at two universities and three different educational programs. The experiment was carried out within courses in property valuation, at the respective university, where both courses use the same textbook on property valuation. The design of the experiment is influenced and constructed in line with the experiment by Palm and Andersson (2021). The design is also inspired by Northcraft and Neale (1987) and Seiler (2014) regarding experimental studies including students. The experiments took place during the final written examination of the course. The examination setting is a classic sit-in exam. The exam consists of a set of questions and one major question consisting of the experiment. The task for the students was to appraise the market value of a property, using the comparable sales method. They were provided with the following supporting materials:

1. Descriptive information about the property,

2. Extract from the Swedish land registration authority,

3. A list of previous sales in the nearby geographical area,

The information from the land registration authority is in the standard Swedish format, which the students had worked with during the course. Information from the land registration authority consisted of address, geographical location, title deed, taxation, construction year, size of building, plot size.

The list of previous sales included information about address title deed, construction year, size of building, plot size, price and date of the sale, a quota measuring selling price/taxation value, and price per square meter. The list of previous sales (extracted from (UCBV ^[1]) consisted of a total number of 87 properties sold during the past year in the same geographical area as the property to be appraised. The student was given the instructions that a client had contacted him or her to conduct an appraisal of a property. Within the instructions, there was information that the client's own appraisal of the value of the property. The information given in the experiment differs between two groups: a low anchor group and a high anchor group. The low anchor group got the information that the client thought of a value of 4.000.000 SEK and the high anchor group got exactly the same information, except for the value, which instead was 6.000.000 SEK.

The instructions in the exam question were divided into three steps; firstly, sort the previous sales list and motivate your sample, secondly, provide three key indicators to be used in the appraisal, and lastly, conduct and appraise the market value of the property. This kind of question and material is something that the students are familiar with from both lectures and exercises.

The fact that this was a part of a final written exam should mitigate the potential for students to engage in heuristic processing solely to minimize cognitive effort, as concluded by Chen and Chaiken (1999)and tested by Palm and Andersson (2021). Regarding ethical considerations for conducting the experiment with students, no approval was required. This was because we utilized the outcome of a written exam and no personal data from the students was collected. Furthermore, the design of the exam question and the objectives of this study did not interfere or affect the grading of the students' answer. The study did not increase the stressfulness of the exam for students, since the complexity introduced by the anchor was a level of complexity that could have been part of the exam anyway. The sole alteration to the exam resulting from the study was the differentiation of “nonsense information” between the two student groups. Specifically, one group received a high amount of this “nonsense information,” and the other group received a low amount.

The experiment was carried out at the final exam of courses in property valuation at the two universities. The students chosen for the experiment came from three different educational programs, Civil engineering in surveying, Real estate brokerage, and Property management. The first group of students, Civil engineering in surveying, have an engineering background and they were 40 in total that answered the question in the exam. The second group of students, Real estate brokerage, have a social science background and they were 87 in total that answered the question in the exam. The third group of students, Property management, have a social science background and they were 46 in total that answered the question in the exam. The second and third groups of students took the same course.

For all the educational programs, two different sets of data were produced:

• Set 1 Property is given a low anchor (4.000.000 SEK)

• Set 2 Property is given a high anchor (6.000.000 SEK)

The two data sets were distributed to the students during their final exam of the course. The task for the participants was to sort the data material using the comparable sales method to appraise a market value for the property. The assignment required them to first sort the historical sales and select the sales that they think are most comparable. They were then to calculate the average regarding price per square meter for the living area, the total property price and the price divided with tax assessment value. Finally, from these averages and the property at hand, reconciling and motivating, a market value for the property was to be given, and this, along with all the related material, was to be handed in after the exam.

Considering the list of 87 sales, not all is appropriate to consider as comparable to the object to be appraised. Twelve observations had no indication of price and sales date, and could not be used, thereby 75 sales remained. The list also included ten properties on leasehold which made them not comparable and thereby 65 sales remain. Two observations were only land sales without any building and should also be excluded, leaving us with 63 sales remaining. If also considering repeated sales and only include the latest of the two, we only had 43 sales left, doing this gives an average price of 4.635′ SEK with the range of 3.000′ – 7.750′ SEK. The students should also consider the building year and the size of the house. Including only houses built 10 years before and 10 years after the property to be appraised left 13 sales. Excluding the smallest and largest houses only allowing houses in the range of plus minus 15 square meters, 4 sales remained. The price range was then 4.620′ to 5.500′ SEK with an average of 5.065′, and an average price per square meter of 43.306 SEK. Applying the price per square meter gives a value indication of 5.153′ SEK when applied on the valuation object. Furthermore, applying the average price divided by tax assessment value gives a value of 4.752′ SEK. The expected value should then be somewhere between 4.752′ and 5.153′ SEK according to the key figures within the comparable sales method.

The result section is divided into two parts for each of the three educational programs. First a descriptive statistic, and second, an ANOVA for each educational program.

The participating students have an engineering background. The total number of participating students writing the exam was 40, and they all successfully provided an appraisal using the comparable sales method.

Table 1 indicates that there is a small difference in the mean appraised value, with the students provided with a high anchor appraising a slightly higher value.

If to illustrate the result in a whisker diagram, the outcome would be as shown in Figure 1.

In the whisker diagram in Figure 1, the difference in appraised values between the group provided with a low anchor and the group with a high anchor is displayed. It provides us with an illustration of how the spread of appraised values from the group provided with a high anchor is wider than for the group with a low anchor. But the difference in appraised value is hard to determine if it is consistent or not.

To investigate the statistical significance of the observed difference, we performed an ANOVA-test to check for the difference between the two groups' mean value, displayed in table Table 2.

From the ANOVA-test, conducted for the civil engineering students, it can be observed that it did not test significantly. The result indicates that there is no statistically significant difference between the two groups.

The participating students have a social science background. The total number of participating students writing the exam was 87, and they all successfully provided an appraisal using the comparable sales method.

Table 3 indicates that there is a difference in the mean appraised value, where the students provided with a high anchor appraises a higher value.

If to illustrate the result in a whisker diagram, the outcome would be as shown in Figure 2.

In addition to the descriptive statistics in Table 3, Figure 2 indicates that there is a clear difference in the mean appraised value between the two groups.

To investigate the statistical significance of the observed difference, we performed an ANOVA-test to check for the difference between the two groups' mean value, displayed in Table 4.

From the ANOVA-test, conducted for the real estate brokerage students, it can be observed that the F-value is 33,024 and significant. The result indicates that there is a statistically significant difference between the two groups, and thus we accept the hypothesis of an anchor effect within this group of students.

The participating students have a social science background. The total number of participating students writing the exam was 46, and they all successfully provided an appraisal using the comparable sales method.

Table 5 indicates that there is a difference in the mean appraised value, where the students provided with a high anchor appraises a higher value.

If to illustrate the result in a whisker diagram, the outcome would be as shown in Figure 3.

In addition to the descriptive statistics in Table 4, Figure 3 indicates that there is a clear difference in the mean appraised value between the two groups. We can also see that the spread within the two groups is less than for the groups in the two previous (Surveying and real estate brokerage) tests.

To investigate the statistical significance of the observed difference, we performed an ANOVA-test to check for the difference between the two groups' mean value, displayed in Table 6.

From the ANOVA-test, conducted for the Property management students, it can be observed that the F-value is 48,935 and significant. The result indicates that there is a statistically significant difference between the two groups, and thus we accept the hypothesis of an anchor effect within this group of students.

To test how the anchor effect works between groups with different educational backgrounds such as civil engineering and social sciences, a between-group experiment was conducted through Scheffe's multiple comparison (Full table in Appendix 1).

As seen in Table 7, the only comparison that is not tested significant (on 5%) is the test between the group of students with an engineering background with a low or high anchor. For the rest of the pair of groups, we can most certainly conclude that there are differences between the appraised values provided by the students, as the test have significant results.

The result of Scheffe's multiple comparison analysis also adds to the acceptance of H2 and leads to the conclusion that coming from an engineering background mitigates the risk for anchor effects in the appraisal process.

The results of this study confirm that anchoring is a robust cognitive bias in property valuation, even among students with formal training in property valuation. The statistically significant differences in appraised values between low and high anchor groups among students with a social science background support Hypothesis 1 (H1), indicating that anchoring effects are present and measurable. By that it adds to the findings in the studies of George et al. (2000), Northcraft and Neale (1987),and Palm and Andersson (2021).

Furthermore, the absence of statistically significant differences among students with an engineering background suggests that their educational framing may mitigate susceptibility to anchoring. This supports Hypothesis 2 (H2), which posits that students with an engineering background are less prone to anchoring bias. These findings align with Bellman et al. (2016), who argue that engineering education fosters more standardized and model-driven reasoning, potentially reducing reliance on cognitive shortcuts.

The implications of these findings are twofold. First, students with different educational backgrounds are influenced by cognitive biases to a greater or lesser degree. This suggests the importance of considering educational background when designing valuation training. Second, engineering students who are trained in mathematical models and are used to applying model thinking suffered less from cognitive bias in our experiment. Palm and Andersson (2021) suggested that awareness alone is not sufficient to avoid behavioral biases, but that it also takes training. Following up on those findings, our results suggest that training in model thinking may be an alternative or supplementary route to reduce the influence of cognitive biases.

Aligning with the persistent anchoring effects observed in previous studies (Palm and Andersson, 2021; Waszczuk, 2024; Węgrzyn and Kuta, 2024), this study contributes to the growing body of literature advocating for enhanced behavioral training in valuation curricula. The result presented here supplements the existing literature and suggest that also training in model thinking is a viable way to reduce the risk of cognitive biases.

The findings of this study, in combination with previous research, have practical implications for real estate education and professional development. Educational institutions should consider integrating behavioral finance and cognitive bias awareness into valuation curricula. This includes not only teaching valuation techniques but also, and most importantly, fostering critical thinking and decision-making skills to help future appraisers recognize and mitigate anchoring effects. Teaching alone, without training, is not sufficient, as previously shown by Andersson and Palm (2021).

Not only the educational institutions but also the professional bodies may benefit from incorporating training that address cognitive biases in valuation, thereby enhancing the robustness and reliability of appraisal outcomes. Furthermore, the results from this study suggests that structured thinking within clearly defined models might mitigate the problem of cognitive bias, as the results show that engineering students are less prone to falling victim to such bias than social science students are.

Our recommendation for both educational institutions and professional bodies is to combine training in behavioral economics with a well-structured model thinking.

The results are also of interest from a broader societal perspective. With the rapidly growing access to AI-tools for property valuation, the supply of cheap and fast automatized valuations is abundant. Considering our results, such valuations risk affecting the thinking of those who need property valuation, such as banks or households. This implies that the professional body of valuers, credit institutions and the governing authorities need to consider the wider consequences of those risks, such as possible systemic risks.

Future research should explore how specific pedagogical interventions can further mitigate anchoring and other cognitive biases in professional practice, perhaps combining training in behavioral economics with training in model thinking.

One other implication for further studies could be how students' merit points may influence the anchor effect. One could argue that students with high scores should be more careful and thoughtful when appraising market values and therefore not as inclined to anchor. In our study, there was a difference between the different student groups, and we cannot rule out that it influenced the result.

For future research, one possibility to enhance the understanding of anchoring effects in property valuation could be to conduct interview studies regarding how the final judgment of market value is concluded, by the professional as well as by the layman.

Although our findings suggest that training in model thinking might reduce the risk of cognitive biases, it should be investigated in future research whether a strong commitment to model thinking at the same time risks lead the valuer to neglect relevant soft information and to not fully apply the sound judgment that (Appraisal institute, 2020) emphasizes.

The supplementary material for this article can be found online.