Edelweiss Applied Science and Technology (ISSN: 2576-8484)

Introduction-The Importance of Corporate Valuation as a Topic in Finance 

An excursion through the economics and finance literatures shows a deep focus on the valuation of a company. The company is usually valued along a series of economic indicators, involving the company and the market. The soft, human factor is also considered, such as the perception of how the company treats its employees, how the company pays attention to the environment, and so forth [1-3]. One need only listen and look at the daily news to see the clear evidence of companies trying to optimize their price by appealing both to the rational economics of their business and to represent themselves as guardians who uphold the publics values [4-6]. 

The notions of corporate reputation and HR (Human Resources) have become increasingly intertwined in the last decade. Corporate reputation is a function of corporate performance, as well as image of the corporation both in the press and in the social media, respectively. With the increasing prevalence and power of social media, and the rapidity with which negative information can be spread, it becomes increasingly important to understand the way the ordinary person converts messaging about the company to estimated price of a companys shares on the stock market. In recent years, researchers have moved from direct questionnaires to more direct measures. One of these is a simulated auction, where respondents are asked to invest in a new product [7]. Rather than instructing the respondent to rate how she or he likes or would purchase a product, the respondent is given some money, and asked either to bid or to invest. The foundational notion is that homo economicus is a more realistic judge of a product or company than is homo emotionalis. That is, when people are instructed to put up money, whether their own or hypothetical money, they tend to act more realistically than when they are instructed to rate their feelings. A lot of this effort can be found in the developing fields of Behavioral Economics and Behavioral Finance, which look at the softer, human side of what was once considered simply as a rational set of behaviors and decisions [8-10]. In previous studies, author Moskowitz presented an approach called Mind Genomics [11,12]. 

Mind Genomics is an emerging psychological science whose goal is to understand how people make decisions under uncertainty. Mind Genomics works by simple experiments, presenting the respondent with combinations of elements or ideas about a situation, e.g., a company, with these combinations created according to a systematic plan known as an experimental design. The respondent evaluates a set of these, which the messages repeating, but in different combinations. The experimental design ensures that the different elements are statistically independent of each other. The data generated by the respondent are analyzed by regression analysis to reveal how each message drives the rating. When Mind Genomics combines with economic aspects, such as price willing to pay, we enter a new world of what might be called Cognitive Economics. Rather than instructing the respondent to evaluate the test stimuli, the combinations or vignettes, in terms of emotion, we instruct the respondent to act as a more rational measuring instrument. The respondent is instructed to read a test vignette, i.e., a concept presenting information about the company, and then rate the company on an economic scale, in this study specifically the expected dollar price of a share. Observations from unpublished data suggest that when respondents rate vignettes on economic dimensions, specifically price related, they appear to be conservative. People may be interested in an idea and willing use extremes to express their feelings, i.e., that they love or hate the idea. When the rating becomes dollars, an economic variable, respondents are not loose and free with their economics-based evaluation. 

Constrained, and there are fewer swings in the magnitude of the rating, even among those who are segmented by the patterns of what they like (so-called mind-sets). That is, people may love or hate what is represented in a vignette, but their ratings of the monetary value will be much narrower, and more logical. The Mind Genomics process and its application to behavior economics/behavioral finance. The origin of this study was the desire to understand how people respond to news stories about corporate governance and the behavior of employees. The study applies Mind Genomics to assess the finance-oriented response to messages about a corporation. The study shows the estimated dollar value of a share of company stoke that can be linked back to each of 16 different aspects of the messages, ranging from the role of the employee giving the information to the way the company demonstrates its values. 

Mind Genomics combines Socratic methods with consumer research. The objective is to understand how people weight the different sources of information when choosing among alternatives. Rather than evaluating the stimulus in splendid isolation, Mind Genomics exposed people to combinations of messages about a topic and deconstructs the responses to these combinations into the contributions of the individual messages. The paradigm simulates the natural experience of individuals, first by showing combinations, and second by presenting these combinations quickly, obtaining the responses quickly, thus forestalling intellectualization of the problem. In the language of Nobel Laurate Daniel Kahneman, when writing about decision making, Mind Genomics employs System 1 thinking, the almost automatic, seemingly unthinking responses to the stimuli of everyday [12,13]. The steps below trace the implementation of the Mind Genomics experiments from problem definition to raw material, test stimulus, analysis, discovery of mind-types, and finally the application of the findings to understanding the larger population beyond the confines of the foundational study. 

Step 1: Define the topic: The topic for Mind Genomics must be limited to something that can be explored in a simple study. The easiest version of Mind Genomics is afforded by the do-it-yourself program, BimiLeap (www.BimiLeap.com). BimiLeap works with 16 different elements, or statements/messages, related to a single topic, with each message painting a word picture appropriate to a topic. The topic is how different messages about a company drive the price a person would pay for a single share of stock. In simple terms, when a company talks about itself, either directly or through its employees, or through advertising and PR, what are the messages which might drive up the value of the companys shares, when the respondent is selected to an ordinary individual with moderate or higher annual income (> $ 50,000)?. 

Step 2: Define four questions about the topic which tell a story: With the Socratic Method, the understanding comes from the pattern of answers to questions. The second step requires that the research ask four questions which revolve around the topic chosen in Step 1. Most researchers are accustomed to asking questions, but not accustomed to the discipline of a defined sequence of questions to elucidate a topic. Mind Genomics requires this laser-focus. Mind Genomics creates a library of integrated knowledge about the topic from the point of view of the human being. It is only by asking a set of related questions about a topic that one can move from the utterly general or the utterly granular to a structure which is almost systematic, archival, and integrated. Table 1 presents the four questions, followed by four phrases or answer for each question. It is clear from the questions that the study is dealing with the description of the individual, the companys behavior, and the customers behavior. These are the aspects which together tell one facet of the story of a company, as narrated by an individual working in the company. 

Table 1: The four questions and the four answers to each question.

 Step 3: Provide four answers to each question or a total of 16 answers: It will be the answers which, in combination, constitute the test stimuli. Table 1 also shows the 16 answers. One of the most frequently asked questions revolves around the desire that there be right and wrong answers. The reality is that in these Mind Genomics explorations there are no right or wrong answers, but only different factoids to which a person responds. It is from the pattern of responses to these elements or answers, responses obtained in short, inexpensive, iterative experiment, that one understands the mind of a person. The questions and the answers in Table 1 are simply first approximations. They will suggest patterns, but they can always be improved, and re-tested, with some questions and answers eliminated new questions and answer inserted, and the process repeated. At some point the relevant patterns will emerge clearly. 

Step 4: Mix the answers (elements) into small, easy to read combinations, so-called vignettes: Each respondent evaluates a unique set of 24 vignettes. Each vignette comprises 2-4 answers, at most one answer from each question, but for a some vignettes the structure deliberately creates incomplete vignettes, with some questions not providing answers for the vignette. This incomplete design structure allows for the model, the equation, to generate coefficients which have absolute, i.e., ratio-scale properties, with a 2 being twice as much as a 1 and half as much as a 4. The rationale for the individual-level experimental design is that in the analysis it will be important to create individual models or equations relating the rating to the presence/absence of the 16 elements. When the combinations are created by experimental design at the respondent level one can be sure of being able to create these individual equations. The rationale for the uniqueness of combinations for each respondent is the desire to cover the space of possible combinations. In most studies using experiment design, the researcher presents the same set of combinations to many respondents, so that the noise is averaged out. Mind Genomics goes in a different direction, assessing many different combinations or vignettes, as many as 1200 for 50 respondents. The noise or extraneous variation is cancelled out, not so much by noise-cancelling repetition, but rather because the pattern emerges more clearly when the test stimuli cover most of the design space. Metaphorically, Mind Genomics can be likened to the MRI of the mind, which creates an image by combining pictures taken from different angles, i.e., from vignettes created from many different combinations. 

Step 5: Create an orientation paragraph to the topic, and an appropriate rating scale: The orientation paragraph should be as general as possible in order not to bias the respondent. Here are the orientation paragraph and the rating scale. Both are minimal in terms of text, yet have a time frame, and a sense of economic reality. Please read the entire screen. It describes a company going into the stock market in 6 months. 
Please tell us what you think the price will be. 1=$ 1/share…9=$ 100/share. 
Rate the price per share 1=1 $/share… 9=$ 100/share. 

Mechanics-setting up the Mind Genomics study

The Mind Genomics process is straightforward to set up, and is aided by a user-friend app, BimiLeap (www.BimiLeap.com). The app guides the researcher towards creating the questions and the answers, as shown in Figure 1, on the left and middle panels, respectively.

Figure 1: The set up for the study. The figure shows the requirement to ask four questions (left panel), the requirement to provide four answers to each question (middle panel), and an example of a vignette combining three answers or elements (right panel)

BimiLeap creates 24 combinations of vignettes for each respondent. Each vignette or combination comprises 2-4 elements or answers, at most one answer from each question. The basic experimental design ensures that all 16 elements are statistically independent of each other. Each respondent evaluates an individual permuted experimental design, with the property that the vignettes or combinations evaluated by each respondent are generally different from the vignettes evaluated by every other respondent [14]. The structure of the design remains the same, but the combinations change. Figure 1 (right panel) shows an example of a vignette as the respondent would see the vignette on a smartphone. The respondent receives an email link from the on-line panel provider (Luc.id, Inc.), with the panel provider specializing in on-line surveys and experiments. The respondents, already members of the 20+ million panels from Luc.id, open the invitation, and if interested, begin the study. 

The typical study lasts approximately 3-5 minutes and is usually done with the respondent paying modest attention to the task. The fact that the respondent is not fully engaged in the study, but rather just does it, is important to the success of Mind Genomics. It means that the respondent is using system 1 thinking, which is automatic, and represents the typical way people think about every day, unimportant issue. Considered opinions, system 2 thinking, may force the respondent into thinking in a rational, considered way, atypical for everyday life [13]. The respondent begins by answer three questions about gender, age, and a third question dealing with interest in the news stories about corporations. The objective of self-profiling is to understand how the respondent describes himself or herself, and she or he approaches the topic of news stories about corporations. The deeper information will come from the pattern of responses, not from the classification.

Analysis

Mind Genomics studies follow a simple structure for analysis. The experimental design enables the creation of models relating the presence/absence of the 16 elements to the dollar value that the respondent thinks that a single share should command. To estimate the contributions of each element to the dollar value of a share, and thus to the perceived value of the element, we use the method of OLS, ordinary least-squares regression, also known as curve fitting. The analysis uses the full set of data for a key subgroup, e.g., gender, to estimate the coefficients of the following equation: Estimated Dollar Value=k1 (A1) + k2 (A2) … k16 (D4). The equation states simply that the estimated dollar value of one share of stock in the corporation, as estimated by the respondent, is a weighted sum of the dollar values of each element, respectively. The higher the coefficient, the great the dollar contribution. When we look at the equation above, we see immediately that the equation lacks an additive constant (intercept), k0, which corresponds to the dollar value of a share in the absence of elements. The additive constant is meaningful when we talk about feelings, such as the likelihood of buying the stock. In that case, the additive constant tells us the basic interest in the stock, based upon the copy. When we deal with the actual price that would be paid, the additive constant has less meaning. To establish the statistical significance of the coefficient, we follow a two-step procedure, based on estimating the equations both with an additive constant, and without an additive constant, respectively. The two steps allow us to calibrate the coefficients and estimate statistical significance. The first step involves estimating the coefficients with an additive constant and without an additive constant, respectively, for the same data set, here the 1200 observations from the total panel. 

Table 2 shows the data for the 16 coefficients. The additive constant is irrelevant here. We focus only on one the coefficients emerging from the two methods of estimating, with an additive constant (intercept), and without an additive constant (through the origin). The two sets of 16 coefficients virtually perfectly co-vary. Figure 2 (left panel) shows for these data how the T statistic varies with the value of the coefficient. We look for a T statistic of 2.0 to define a significant level. The left panel suggests a coefficient of +5 for the equation or model with an additive constant. 

In turn, the right panel of Figure 2 shows that a coefficient of +5 for an equation with an additive constant (our threshold of significance), corresponds to a coefficient of approximately +20. Thus, we assume that the coefficient of +20 or higher for an equation or model without an additive constant, i.e., without an intercept, a model which goes through the origin. Table 3 shows the coefficient for six groups, based upon who the respondents say they are. The groups are Total, Gender, and Age. To make the extraction of key messages easier, Table 3 shows the strong coefficients in bold, and in shaded cells. Strong is defined as a dollar value for the element of $ 21 or higher. 

Table 2: Statistical parameters of the equations relating the presence/absence of the 16 elements to the ratings (dollars would pay). The pairs of columns show the statistical parameter of the equation for each element, first for the model through the origin (no intercept), and then with an intercept (additive constant). The 16 elements are sorted in order of magnitude. 

Figure 2: Scatterplot showing the relation between the coefficient and the T-statistical (left panel, model with an additive constant, i.e., intercept), and a scatterplot showing the relation between the two sets of coefficients estimated for the total panel data of this study (right panel). 

Although the respondents did not estimate the dollar value of each message as a contributor to likely stock price, the regression model enables the estimation of the dollar value. The use of inferential statistics to determine significance is less relevant than a holistic view of the patterns separately strong performing elements driving high dollar values from weak performing elements driving low dollar values. There is very little in terms of general population. It may well be that no elements really drive estimated price willing to pay, or, as other observations suggested, there are differences in price traceable to how people think of themselves, not traceable to who they are suggests the importance of one element only across groups: Customers: give high reviews and ratings. 

How the individual giving the information drives the estimated stock price

The Mind Genomics experimental design permutes the different designs to create 50 different sets of 24 vignettes, most of which differ from each other. It is thus possible to isolate vignettes which the same presenter, defined as the person doing the talking. These are the vignettes whose first element begins with the words MY JOB. 

Table 3: Models relating the presence/absence of elements to the dollar value of the stock. The models are based on WHO the respondent says she/he IS. The table shows strong performing elements ($21 or higher), in shaded cells, and in bold font. 

By sorting the vignettes into the five groups, one group for each of the five answers to Question A, one comes up with a set of homogeneous vignettes in terms of who is presenting the information. The presenter may be either no person (A0), a person who makes the products (A1), a person who sells the product (A2), a person who trains (A3), or a person who manages project timelines (A4), respectively. When we build a model for each of the five strata, the question immediately becomes whether the presenter of the information can drive the price of the other elements or messages. All other factors held constant, does the who does the saying affect the estimated price for a share of stock? 

Table 4 shows the analysis of the data, comprising five equations, one for each of the five groups defined by defined by WHO presents the information. There are only 12 coefficients, one for each of the remaining elements. The four elements corresponding to WHO present the information are no longer part of the model because they are held constant for each model. For these vignettes we raise the criterion of an important element to a slightly higher dollar value of $26 or more for one share to simplify the table. Depending upon who does the presenting, the other elements show variation in the dollar values they command, suggesting that the presenter acts as a director or anchor.

No Defined Presenter

Company: acts legally with its self-interest in mind

Company: takes advantage of social media

Note: Providing news without identifying who said it maintains estimated share value. Providing an employee as source of information decreases estimated share value.

Those who MAKE the products

Customers: give high reviews and ratings

Me: talk honestly about my company in a positive way

Those who SELL the products

Customers: give high reviews and ratings

Customers: love to contact customer service with feedback whether it is positive or negative

Customers: after purchase become evangelists for the products/services

Those who TRAIN within the company

Customers: give high reviews and ratings

Customers: love to contact customer service with feedback whether it is positive or negative

Customers: after purchase become evangelists for the products/services

Those who MANAGE timelines

Me: dutifully follow company policy 

Table 4: Dollar value of elements, with the vignettes stratified by WHO presents the information. Those dollar values of 26$ or higher are shown in shaded cells with bold font. 

The dollar value based upon the interest in stories about companies and their employees/customers

At the start of the experiment, the respondent was asked to select one of three descriptions about interest in how companies treat their customers and how their employees feel. There were four answers, collapsed to three by combining those who selected not interested with those who selected not applicable. Table 5 presents the dollar values for the total panel and for the three groups defined by how they feel about these kinds of stories. The key results from are that there only five relevant elements which drive a contribution to the estimated stock price, with the one element extraordinarily important for the stock price, Customers: give high reviews and ratings Table 5. The other elements add to the value of the stock, but not in a consistent way across all respondents. The key here is reviews and ratings from customers

·       Me: dutifully follow company policy

·       Customers: after purchase become evangelists for the products/services

·       Company: acts legally with its self-interest in mind

·       Me: talk honestly about my company in a positive way

It is important to note that in the Mind Genomics experiment, the test vignettes are presented and respondent to so quickly that the strong performing elements emerge almost automatically, without much conscious processing. 

Mind-Sets, people who think alike with respect to the messages which drive value

One of the major tenets of Mind Genomics is that in every topic, there are different groups of ideas which flow together. These are ideas, not people, but it is only by studying the response of actual people to these ideas is it possible to discover these different groups of ideas, or so-called, metaphorically, mind-genomes.

Table 5: Models relating the presence/absence of elements to the dollar value of the stock. The models are based on what type of story interests the respondent. The table shows strong performing elements (dollar value 21 or higher) in shaded cells, and in bold font.

In simple terms, one can think of a mind-genome as a set of ideas, the combination of which make sense and gives us a feeling that there is a coherent theme. Moving further into the metaphor, the notion is that there are probably a limited number of mind-genomes for any topic, that these genomes representing different mental primaries of a topic, and that the mind-genomes may be uncovered through the analysis of a persons pattern of responses to a set of relevant stimuli. The mind-genomes are uncovered empirically. They may be hypothesized to exist, based upon ones experience, but the real test is whether these coherent groups of ideas can be demonstrated. The approach to discover the mind-sets uses the statistical method of clustering. The clustering is done on the pattern of coefficients generated by each respondent in order to discover groups or clusters of individuals showing similar patterns of coefficients. The name of the cluster (renamed mind-set) comes from the commonality of the highest scoring coefficients in the cluster. There are two steps requiring judgment. The first is interpretability-does the pattern make sense? The second is parsimony is it possible to develop only a limited number of patterns which cover the range well, albeit not necessarily perfectly? Mind Genomics provides the appropriate data for clustering. 

The inputs are the 50 rows of data, one row for each of the 50 respondents. The columns are the 16 coefficients, each column being the dollar value of the specific element for the specific respondent. It is now, in the clustering phase, that we see the power of the individual-level modeling, made straightforward by the experimental design. Each respondents ratings can be immediately analyzed by OLS regression to estimate the 16 dollar-values for one share of stock, a dollar value for each of the 16 elements. Statisticians have provided many methods for clustering. The clusters, groups of like-objects are not fixed in stone, but rather are constructed according to objective mathematical criteria.

The clusters themselves are simply a collection of objects which fulfill the clustering requirements but may not have meaning when interpreted. It is left to the researcher to decide the clustering method to use and within that method to choose the number of clusters to create. In the end, the action of clustering is more of a heuristic to identify patterns in the data rather than the specification of tight, non-overlapping groups [15]. For the price data, the clustering method, so-called k-means, defines a distance between every pair of respondents based on a definition of distance in quantitative terms. Distance for this study is defined by the dissimilarity of two patterns, one per respondent, based upon the dollar values of the 16 elements. The actual measure of distance is (1-R), where R is the Pearson correlation between the two respondents, using the 16 dollar-values as cases. 

The Pearson R reaches a high of +1 when the two patterns are perfectly related to each. In that case, with R=1, the distance is 0 (1-1=0). The Pearson R reaches a low of -1 when the two patterns are perfectly, but inversely related to each other, so that increases in the dollar value of Respondent A is counterbalanced by the same relative magnitude of decrease for Respondent B. In that case, the distance is 0 (1 - -1 = 2). When the clustering process finishes, it emerges with two, three, or, when necessary, even more groups. These groups are quite different in the patterns. The important thing for Mind Genomics is that these clusters or Mind-Sets are parsimonious (few) and interpretable (tell a story). Table 6 shows the dollar value of the different elements, by total, and then by two mind-sets. 

The mind-sets make sense when we look at the dollar values. The clusters or mind-sets are not dramatic. Often, a different rating scale calling into play emotion instead of price generates more dramatically different mind-sets. The one overwhelmingly strong element is D1 (Customers: give high reviews and ratings.) That strong performing element may reflect the increasing use of customer feedback mechanism as a business tool, both for improving the business and for communicating the business to those who buy of products and services. Beyond the strongest element there are two clearly different mind-sets. Mind-Set 1 (MS1) assigns high share values to messages about good corporate governance. Mind-Set 2 (MS2) assigns high share values to messages about customer intimacy and involvement. 

Table 6: Dollar values of the elements by total and two emergent mind-sets.