An investigation into the impact of previous expert value estimates on appraisal judgment

Abstract. Is appraisal judgment influenced by the previous value estimates of other experts? This paper documents an investigation into this question. The literatures on appraisal smoothing and heuristic anchoring were explored to build a theoretical and empirical base. Research methods involved asking apprentice and expert appraisers, some supplied with the previous value estimate of an anonymous expert, to estimate the value of a vacant tract of industrial land. The strong support expected for the contention that appraisers are influenced by the previous value judgments of anonymous experts was not found. Whereas differences between the groups supplied with the previous value estimate and those who were not were in the direction consistent with anchoring, these results were not statistically significant, suggesting that anchoring may be more subtle than generally believed.

Introduction

Ibbotson and Siegel (1984) recognized that real estate return series constructed using appraisals as substitutes for transaction information exhibited reduced variability. This problem, labeled appraisal smoothing, was attributed by the authors to inherent weaknesses in the process of real estate valuation, that is, to a reliance on historic cost and transaction information. Cole (1988) was able to demonstrate this reduction in the variability of valuation-based series by comparing them to actual transaction-based indices. According to Cole, several factors may contribute to appraisal smoothing: (1) factoring past value estimates into current appraisal judgments, (2) annual rather than quarterly appraisal cycles, and (3) influence of desired client outcome.

Geltner (1989a) provided a precise definition of appraisal smoothing as the ratio of the standard deviation of true portfolio property values to the standard deviation of appraised portfolio property values. Despite the suggestiveness of the term "appraisal smoothing," this view defines an observed phenomenon but does not attempt to assign behavioral causality. Geltner does allude to a cause of appraisal smoothing as a function of directed valuations and in a subsequent paper (Geltner, 1989b), offers lack of confidence (and hence a reliance on past "acceptable" value estimates) and valuation timing as two other possible explanations of the phenomenon. These candidate causes echo Cole's original list.

Recently the quest to explain reduced variation in appraisal-based series has tended to center on appraisal smoothing as a function of appraisers rationally weighting previous value estimates against subsequent market information. Quan and Quigley (1991) argued that smoothing is the consequence of an updating strategy employed by appraisers, that is, appraisal smoothing does not result from flawed methodology, incompetence or inappropriate influences but rather from the appraiser's rational choice to rely increasingly on previous value estimates in the face of greater market uncertainty. While this view has been endorsed by others (notably Geltner, 1993), its veracity and robustness remain empirically untested.

The use of anchors in problem solving and judgment has been explored and is documented in a rich literature. Slovic and Lichtenstein (1971) identified an anchoringand-adjustment heuristic (or rule-of-thumb) with which humans make value estimates by starting from an initial reference value (anchor) and adjusting from this reference point as evidence is assimilated. These authors suggested that decision bias may occur due to insufficient adjustment from the anchor. Another important historical work in this area is Tversky and Kahneman (1974).

More recent effort has uncovered evidence of anchoring in time series forecasting (Lawrence and O'Connor, 1992); in personal assessments of future efforts and performance although not in actual performance and efforts (Switzer and Sniezek, 1991); in probability assessments (Wright and Anderson, 1989); among students participating in a laboratory experiment involving a simulated economy (Sterman, 1989); and in utility assessments (Johnson and Schkade, 1989). Block and Harper (1991) found anchoring evidence among students estimating familiar and unfamiliar quantities (e.g., number of spokes in a wheel versus number of states voting to ratify the Equal Rights Amendment) but found no relationship between anchoring and overconfidence. The generalizability of anchoring beyond the laboratory has not gone unquestioned. Important challenges have been brought up in Hogarth (1981) and Berkeley and Humphreys (1982).

The anchoring heuristic in real estate pricing decisions was studied in Northcraft and Neale (1987). Here students and real estate sales agents visited a residence currently for sale, were provided a packet of relevant information and were asked to estimate the appraised value. The listing price was provided to subjects but was varied at four different levels, a low listing price, a moderately low listing price, a moderately high listing price, and a high listing price. The investigators found a strong anchoring impact among both student subjects and professional sales agents. Further, the impact did not diminish as the anchor became less credible (low and high listing prices).

Despite the importance of this work, it reveals nothing about expert valuation behavior relevant to the appraisal smoothing question. Real estate sales agents are marketing experts but are not subject to the rigorous valuation training and held to high reporting standards as are the true real estate valuation experts, the appraisers. The Appraisal Institute (1992) prescribes a set of cues and a methodology, called the appraisal process, to be used by appraisers when estimating real estate values. Furthermore, appraisers go through a professional designating process and a state of certifying process which require that they demonstrate significant familiarity and competency with this normative process. It is important to note that previous valuation judgments enjoy no role in the prescribed appraisal process.

Overwhelmingly the studies that have revealed heuristic anchoring have focused on the behavior of students or other novices. These include the Northcraft and Neale investigation whose real estate sales agent subjects would typically lack the normative training necessary to elevate their valuation behavior to the expert-like. An exception to the use of novice subjects is found in the behavioral auditing literature of accounting where some expert auditors have been used as subjects but where results have been mixed and findings unclear. (See Shanteau, 1989, for a discussion of this literature.) Anchoring behavior and its attendant potential for bias remain well documented among novice problem solvers but generally unexplored among experts.

The extent and nature of expert appraiser anchoring is unclear but must be understood before reliance-upon-previous-value-judgments can be accepted or rejected as a cause of appraisal smoothing. This paper documents the initiation of an investigation into this issue. Because of the strong evidence of notice anchoring and because of the theoretical postulate of Quan and Quigley, Geltner and others, the research hypothesis driving the investigation is that in the face of market ambiguity and uncertainty, the behavior of expert real estate appraisers will be influenced by previous value judgments. Research methods designed to test this hypothesis are discussed next.

Research Methods

Because relevant data that are appropriate for the traditional, regression-based methodology do not exist, a controlled experiment was designed to investigate the hypothesis that expert appraisers will demonstrate anchoring behavior. This methodology is a powerful tool enjoying the faith of such important research disciplines as psychology and medicine. It allows a tight focus on the impact of a few independent variables on the response variables. Since extraneous variation is reduced or controlled, the impact of independent variables can readily show itself. Under such conditions, small sample sizes of thirty or less will demonstrate significant statistical power.

A two-factor experimental design was employed, the two factors being (1) anchor, and (2) level of expertise. The anchor variable was fixed at two levels, anchor versus no anchor. The existence of an anchor was operationalized by providing experimental subjects with the previous value estimate of an anonymous expert. No such judgment was provided in the no anchor case. Level of expertise was fixed at two levels, appraiser apprentice and expert so that any dampening impact of expertise on anchoring could be investigated. An appraiser apprentice was defined as an individual who had completed appraisal course work and who had worked for no more than five years in an appraisal organization as an appraiser trainee under the supervision of an expert appraiser. Expert appraisers were defined as practicing real estate appraisers who had attained the highest designation of the premier real estate appraisal organization (MAI of the Appraisal Institute).

A list of potential study participants from the Atlanta area was constructed from the membership roster of the Appraisal Institute. MAIs from this list were selected at random and were asked to participate in the experiment. No MAI contacted refused to participate. Participating MAIs were also asked if apprentice appraisers within their organizations could be solicited for participation. Again all agreed, and all apprentice appraisers asked to participate did so. Seventeen expert appraisers (MAIs) from the Atlanta area were also solicited from a continuing education class offered by the Atlanta Area Chapter of the Appraisal Institute.

The experimental task involved the valuation of a vacant parcel of industrial land located in the northern suburbs of Atlanta and was designed to be representative of real world appraisal assignments. The case was developed from an actual report provided by an Atlanta area expert. Factual data from the report were extracted, revised to meet the design of the experiment and incorporated into the case. Data items included in the case were identification of the subject, purpose of the appraisal, neighborhood data, neighborhood map, property data, subject plate, subject photographs, five comparable sales (including photographs), and a sales map. Comparable sales data consisted of contrived transaction information on five vacant parcels of land exhibiting significant similarity to the subject tract being valued. The transaction information was manipulated so that no obvious pattern (in terms of size of tract, date of sale, etc.) existed. The highest and lowest per acre sale prices were assigned to the two tracts designed to be most similar to the subject being appraised. This variability, ambiguity and lack of transaction price pattern was judged to maximize the probability of evoking anchoring behavior in experimental subjects by introducing significant market uncertainty into the appraisal task. On a per-acre basis, the comparable transaction information is provided in Exhibit 1.

Each experimental case also included a problem statement. Two different problem statements were used in the experiment, one for the no anchor case, the second for the anchor case. The two problem statements specified the same appraisal problem, same date of valuation and same instructions for completing the experimental task. The problem statements differed in their characterization of the information provided in the case, the no anchor case stating, "Enclosed you will find the data and information which resulted from a diligent search of the market," and the anchor case stating, "Enclosed you will find excerpts from an appraisal report recently prepared on the subject by a local MAI." Following the problem statement for each case was a blank work sheet and for the anchor case only, following the work sheet but preceding the actual case information, was a letter of transmittal that specified the value estimate of the local MAI (i.e., the anchor), of $88,000 per acre. All other items were identical for the anchor case versus the no anchor case.

Three expert appraisers not used as experimental subjects verified the real world representativeness of the case. A trial run involving the participation of three appraiser trainees and eight Georgia State University undergraduate real estate students resulted in minor revisions to the problem statements and to one sale.

Experimental sessions began on August 18, 1993, and the last one occurred October 22, 1993. Each participating subject received at random one of the experimental cases (anchor versus no anchor), and was requested to estimate the market value of the described property based only on the information provided. To minimize deviation in the response variable (market value estimate) due to extraneous factors, strict quality control was observed. The same written instructions and the same administrator were employed throughout the experiment. All verbal communications between experimenter (administrator) and subjects were minimized. Settings were either the normal workplace of subjects or classroom environments. Twenty-eight apprentices and thirty experts participated in the study. One-half of the subjects from each expertise group was randomly assigned the anchor case while the remaining half was randomly assigned the no anchor case.

Any anchoring behavior should be manifested in its impact on group location. The experiment was designed so that all comparable sale properties exhibited significant similarity among themselves and the subject being valued. Further, all price pattern was removed from the data. In such a situation the group central tendency value judgment should be similar to the central tendency of the comparable sales data (i.e., $85,061 per acre mean and $85,001 per acre median) as long as there is no anchor impact. If an anchor effect exists, the previous value judgment of $88,000 per acre should exert a positive pull on the central tendency of experimental groups exposed to the anchor. Therefore, under conditions of an anchor impact (consistent with the research hypothesis),

(mu)^ sub A>(mu) ^ sub NA, (1)

where

(mu) ^ sub A^ = the central tendency of the value judgments of the anchor group; (mu) ^ sub NA^ = the central tendency of the value judgments of the no anchor group. The standard rules for the development of test hypotheses from the research hypothesis (see, for example, Daniel and Terrell, 1992, pp. 308, 309) state that (1) the null hypothesis must include a statement of equality (equal to, less than or equal to, greater than or equal to), and (2) the research hypothesis (what is expected) should go into the alternative hypothesis provided that the first rule is not violated. Applying these rules to the research hypothesis (equation 1 above), yields the following test hypotheses:

H ^ sub o^: (mu)^ sub A^

H ^ sub A^: (mu) ^ sub A^>(mu) ^ sub NA. (3)

The evidence upon which these test hypotheses were evaluated is discussed in the following section.

Results

Results of the experimental sessions are summarized in Exhibit 2. A perusal of these results suggests no support for equation 1 and therefore no support for the research hypothesis. Statistical tests confirm this observation. Parametric t-tests for the difference between the means of two independent samples reveal no evidence to reject the null hypothesis of no difference in anchor versus no anchor group means for either the apprentice or expert groups (one-tailed p-values of .481 and .238, respectively). Since the parametric properties of the populations from which the samples were taken are unknown, nonparametric median tests were also performed on each of the two expertise groups. Again no significant differences were found with p-values of .2206 and .9319, respectively, for the apprentice and expert groups. Simply stated, the differences between anchor and no anchor group means are in the direction consistent with anchoring, but these differences are statistically trivial.

Whenever no significance is discovered in data, the concern must be for the probability of committing a Type II error, that is, with the failure to detect within the data collected a significant difference that exists in the populations of interest. The a posteriori probabilities of committing Type II errors were calculated for each test employed using the actual data collected, fixing Type I error probabilities at 5% and varying the anchor group population mean from $85,500 to $88,000 per acre while assuming the no anchor group population mean was equal to $85,000 per acre. (Nonparametric probabilities were based on asymptomatic relative efficiency of .955 as reported in Gibbons, 1976.)

These probabilities are offered in Exhibit 3A-D and indicate that for the expert group the ability to detect strong anchoring is quite good. Type II error probabilities for the expert group range from about 2% for a hypothesized anchor group mean of $88,000 to around 10% at $87,400. At an anchor group mean of $86,300 the probability is roughly 50% which indicates that the existence of weak anchoring is much harder to detect. However the likelihood that any expert subject was influenced by the previous value judgment anchor is remote since as shown in Exhibit 2, no expert subject provided with the anchor estimated the value of the property at more than $85,000 per acre. Because of larger sample variability, the probability of Type II errors is uncomfortably high for the apprentice group (see Exhibit 3C-D) meaning that the ability to detect anchoring among apprentice subjects is not strong.

Conclusions

Despite empirical precedence from novice problem solving and simple problem solving domains, this investigation did not find support for the contention that either apprentice or expert appraisers faced with a real estate valuation task are influenced by the previous value judgments of anonymous experts. While group mean differences were consistent with the existence of anchoring, the differences were not statistically significant. The lack of support for strong anchoring is especially strong for experts in light of Type II error probabilities (probability of failing to find significant differences in the data when they exist in the population) although the ability to detect weak anchoring influences within the data is limited. Less convincing is the no anchoring conclusion for apprentice appraisers due to larger sample variability and the resultant higher Type II error probabilities. Nonetheless the view that appraisers will routinely utilize previous value judgments as valuation cues is called into question. The study was made as true to real world settings as possible to protect against threats to generalizability, but absolute extension of findings into real world settings is cautioned.

Beyond deepening our understanding of the descriptive process of appraisal (i.e., how appraisers actually derive value judgments as opposed to how they "should"), these results have implications for research into appraisal smoothing since they are suggestive of what appraisers might do in real world tasks with reference points from sources other than anonymous experts. Especially compelling is the need to investigate the role of appraisers' own previous value estimates in the formation of valuation judgments in an environment of transaction noise, incomplete data or significant variability in comparable data. Currently an investigation to study just this issue is under way.

References

Appraisal Institute, The Appraisal of Real Estate, Chicago: Author, tenth edition 1992. Berkeley, D. and P Humphreys, Structuring Decision Problems and the `Bias Heuristic,' Acta

Psychologica, 1982, 50, 201-52.

Block, R. A. and D. R. Harper, Overconfidence in Estimation: Testing the Anchoring-andAdjustment Hypothesis, Organizational Behavior and Human Decision Processes, 1991, 49, 188-207.

Cole, R., A New Look At Commercial Real Estate Returns, Ph.D. dissertation, University of North Carolina at Chapel Hill, Ann Arbor, Mich.: University Microfilms International, 1988. Daniel, W W. and J. C. Terrell, Business Statistics for Management and Economics, Boston: Houghton Mifflin, sixth edition 1992.

Geltner, D., Estimating Real Estate's Systematic Risk from Aggregate Level Appraisal-Based Returns, AREUEA Journal, 1989, 17:4, 463-81. , Bias in Appraisal-Based Returns, AREUEA Journal, 1989, 17:3, 338-52. Estimating Market Values from Appraised Values without Assuming an Efficient Market, Journal of Real Estate Research, 1993, 8:3, 325-45.

Gibbons, J. D., Nonparametric Methods for Quantitative Analysis. New York: Holt, Rinehart and Winston, 1976.

Hogarth, R., Beyond Discrete Biases: Functional and Dysfunctional Aspects of Judgment Heuristics, Psychological Bulletin, 1981, 90, 197-217.

Ibbotson, R. and L. Siegel, Real Estate Returns: A Comparison with Other Investments, AREUEA Journal, 1984, 12:3, 21942.

Johnson, E. J. and D. A. Schkade, Bias in Utility Assessments: Further Evidence and Explanations, Management Science, 1989, 35, 40624.

Lawrence, M. and M. O'Connor, Exploring Judgmental Forecasting, International Journal of Forecasting, 1992, 8, 15-26.

Northcraft, G. B. and M. A. Neale, Experts, Amateurs, and Real Estate: An Anchoring-andAdjustment Perspective on Property Pricing Decisions, Organizational Behavior and Human Decision Processes, 1987, 39, 84-97.

Quan, D. C. and J. M. Quigley, Price Formation and the Appraisal Function in Real Estate Markets, Journal of Real Estate Finance and Economics, 1991, 4, 127-46.

Shanteau, J., Cognitive Heuristics and Biases in Behavioral Auditing: Review, Comments and Observations, Accounting, Organizations and Society, 1989, 14, 165-77. Slovic, P and S. Lichtenstein, Comparison of Bayesian and Regression Approaches to the Study of Information Processing in Judgement, Organizational Behavior and Human Performance, 1971, 6, 649-744.

Sterman, J. D., Misperceptions of Feedback in Dynamic Decision Making, Organizational Behavior and Human Decision Processes, 1989, 43, 301-35. Switzer, F. S. and J. A. Sniezek, Judgment Processes in Motivation: Anchoring and Adjustment Effects on Judgment and Behavior, Organizational Behavior and Human Decision Processes, 1991, 49, 208-29.

Tversky, A. and D. Kahneman, Judgment Under Uncertainty: Heuristics and Biases, Science, 1974, 185, 1124-31.

Wright, W. F. and U. Anderson, Effects of Situation Familiarity and Financial Incentives on Use of the Anchoring and Adjustment Heuristic for Probability Assessment, Organizational Behavior and Human Decision Processes, 1989, 44, 68-82.

*Department of Real Estate, College of Business Administration, Georgia State University, PO Box 4020, Atlanta, Georgia 30302-4020.

Date Revised-October 1994; Accepted-July 1995.

Copyright American Real Estate Society 1997
Provided by ProQuest Information and Learning Company. All rights Reserved