Essays on behavioural economics and cost overuns

(1)

Essays on behavioural economics and cost overruns

Fredrik Brunes

Doctoral Thesis

Building & Real Estate Economics

Department of Real Estate and Construction Management Royal Institute of Technology

Kungliga Tekniska Högskolan

(2)

Royal Institute of Technology (KTH) Building & Real Estate Economics

Department of Real Estate and Construction Management SE–100 44 Stockholm

Printed by Universitetsservice US-AB Stockholm TRITA-FOB-DT 2015:1

(3)

Abstract

This dissertation consists of five essays, four on cost overruns and one on behavioral economics with a focus on real estate investments. The main aim of the thesis was to answer the questions: • Is it common with cost overruns in infrastructure projects? • What explains the cost overruns? • What can be done to prevent cost overruns? This thesis contributes to the current state of knowledge within the field: • It is yet another evidence that cost overruns are a significant problem. • Optimism bias is an explanation of cost overruns. However, the thesis adds, by Article two, a rationality explanation. • Dismisses in part the Successive method as a tool to reduce cost overruns. • Provide recommendations on how action against cost overruns can be systematized. The first article, which is a result of my Licentiate thesis deals with real estate investments and the fact that investors sometimes are influenced by behavioral aspects that make them deviate from what is rational. The result is somewhat uncertain but shows that investors in these cases were influenced by myopic behaviour. The second article deals with cost overruns and rationality. The question is whether it can be rational in a procurement, using a unit‐price contract, to expect cost overruns. The result shows that it is possible in situations where the decision maker has to take renegotiation costs and monitoring costs into account. The third article examines how well the successive method can pvents cost overruns. Established results in research on cost overruns indicated that strategic factors, psychological factors (optimism bias) and technical factors are important to explain cost overruns. The review suggests that the Successive method has limitations as a method to reduce cost overruns because it cannot deal with strategic and psychological factors. The fourth article focus on why cost overruns incurred. The article presents a new framework based on microeconomic cost theory and also results based on a survey to project managers in infrastructure projects, suggesting that cost overruns are common and that cost overrun mainly are due to the optimism bias. The issue addressed in the fifth article is what can be done to prevent cost overruns. The conclusion, based on questionnaire survey and literature review, suggests that a variety of policy measures are needed at different levels such as (1): Organizational macrostructure where e.g. cost overruns in a project in one region leads to less projects in that region.(2) Organizational quality: improved transparency within organizations to see where and when cost overruns occur. (3) Organizational processes: e.g. the use of external reviewers.

(4)

Sammanfattning

Den här avhandlingen består av fem uppsatser, fyra om kostnadsöverskridanden och en om beteendeekonomi med inriktning på fastighetsinvesteringar. Huvudsyftet i avhandlingen har varit att besvara frågorna:  Är det vanligt med kostnadsöverskridanden i infrastrukturprojekt?  Vad förklarar kostnadsöverskridanden?  Vad kan göras för att förhindra kostnadsöverskridanden? Avhandlingens bidrag till rådande forskningsläge är att  Den är ännu ett belägg för att kostnadsöverskridanden är ett påtagligt problem.  Optimism bias är en förklaring till kostnadsöverskridanden. Men avhandlingen adderar, genom artikel två, möjligheten att det kan vara rationellt att utforma en upphandling så att den leder till kostnadsöverskridande..  Avhandlingen avfärdar delvis den Succesiva Metoden som metod att minska risken för kostnadsöverskridanden..  Slutligen ger avhandlingen rekommendationer hur kostnadsöverskridanden kan åtgärdas genom. Den första artikeln, som är ett resultat av min licentiatavhandling, behandlar fastighetsinvesteringar och det faktum att investerare ibland är påverkade av beteendeaspekter som gör att de systematiskt avviker från vad som är rationellt. Resultatet, som är något osäkert, visar att investerare i de aktuella fallen var influerade av myopiskt tänkande. Den andra artikeln behandlar kostnadsöverskridanden och rationalitet. Frågan är om det kan vara rationellt vid upphandling, enligt fast pris per enhet, att förvänta sig kostnadsöverskridanden. Resultatet visar att det är mycket möjligt vid situationer där beslutsfattaren har att ta hänsyn till omförhandlingskostnader och övervakningskostnader. Den tredje artikeln granskar den Succesiva Metoden som är en metod som används för att förhindra kostnadsöverskridanden. Etablerade resultat inom forskningen om kostnadsöverskridanden pekat på att strategiska faktorer, psykologiska faktorer (optimism bias) och tekniska faktorer är viktiga. Den Successiva Metoden används dels för att upptäcka potentiella risker med infrastrukturprojekt men även för att förhindra kostnadsöverskridanden. Granskningen pekar på att den Successiva Metoden har klara begränsningar för att minska kostnadsöverskridanden eftersom den inte på ett bra sätt förhindrar strategiska och psykologiska faktorer. Den fjärde artikeln frågeställning är hur vanlig kostnadsöverskridanden. Artikeln behandlar varför kostnadsöverskridanden uppstår. Resultatet som baseras på enkätundersökning till projektledare i infrastrukturprojekt, tyder på att kostnadsöverskridanden är vanliga samt att kostnadsöverskridanden främst beror på optimism bias. Artikeln presenterar också en ny struktur för förklaringar som bygger på mikroekonomisk kostnadsteori. Frågeställningen i den femte artikeln är vad som kan göras för att förhindra kostnadsöverskridanden. Slutsatsen, baserad på enkätundersökning och litteraturgenomgång, pekar på att en mängd policyåtgärder behövs på olika nivåer bl a organisatoriska förändringar som straffar de regioner där

(5)

kostnadsöverskridanden uppstår, ökar transparensen i processerna och ökar användningen av externa bedömningar av projekt och kostnadskalkyler.

(6)

Acknowledgements

There are many who I would like to thank for their support in the process of the thesis. This concerns both direct support in the work with the thesis and indirect support through discussions on education, the third task of a University (i.e., meeting with practitioners) and support from friends. Direct support Regarding the thesis I would like to thank Professor Hans Lind whose contribution has been significant. With a never dwindling patience and with lots of ideas, he has helped me in my work. I would especially like to thank Hans for the introduction and guidance in research theory. I would also like to thank Associate Professor Svante Mandell, who was co‐author of the second article of my dissertation. I learned a lot about scientific logic by writing that article. I would also like to thank Professor Mats Wilhelmsson, who gave great feedback on an earlier version of the thesis. For financial support, I would like to thank Formas and the taxpayers! Indirect support Obviously I would like to thank my current and former colleagues at the Department of Real Estate and Construction Management at KTH, with whom I have had many interesting discussions about research and education. Let me mention a few of my peers like Johan Nystrom, Abukar Warsame, Sigrid Katzler, Han Suck Song, Berndt Lundgren, Olof Netzell, Carl William Åström, Anders Hellström, Bo Nordlund, Björn Berggren, Andreas Fili and Bo Söderberg. I would also like to thank Professor Roland Andersson, who supervised me in my master thesis (feels like 20,000 years ago) and recommended me for a job at the department which later led to this thesis. I would also like to thank  Åke Ekwall for many interesting discussions about property issues.  Erik Persson for good advice in teaching and property valuation.  Susanne Hörnfeldt, Johan Böckert, Magnus Stenback and Jonas Petersson, among others, in the real estate sector, that I often have used for reality check.  Many friends / colleagues, through the years, that I have learned a lot from, such as members of Samhällsbyggarna, The Board of Samhällsbyggarna, The Board of SFF, Folkuniversitetet, Skattemyndigheten, KTH Mediproduktion, Studentlitteratur, Fridhems Folkhögskola, Fastighetsnytt, Newton Education, Sveriges Improvisationsteater and persons in private companies such as Newsec, Forum Fastighetsekonomi, and NAI Svefa. Things do not happen by chance ‐ there is a clear logic why I teach and conduct research in Building and Real Estate Economics. The fact that my father, an economist, used to take me along to see every construction site in southern Hälsingland gave me the interest in construction and economics and my mother, a teacher with acting skills, gave me the interest in teaching. Thanks Anders and Christina Brunes and thanks to my sisters, Jenny Brunes and Susanna Jahr, and all relatives that I have a good time with.

(7)

What the heck, I would also like to thank my friends: No, they have not helped me with construction and property issues, but they are good people and good support: Rikard Amcoff, Lars Eriksson, Richard Henriksson, Tobbe Sillén and Pär Viklund. Finally, and most importantly, I would like to thank Jenny Asp and “Plutten” ‐ my two loves ‐ my wife and my son to be born on July 27th. Stockholm 2015‐04‐26

(8)

Part I – Overview

Part II – Articles and general conclusions

Article 1 – Overbuilding in office markets: Are behavioural aspects important? Article 2 – Quantity Choice in Unit Price Contracts Procurements

Article 3 – Successive Principle: will it solve the problem of cost overruns?

Article 4 – Explaining cost overruns in infrastructure projects: A new framework with applications to Sweden

Article 5 – Policies to avoid cost overruns in Infrastructure Projects: Critical evaluation and recommendations

(9)

Part I – Overview

1. Introduction

This thesis deals mainly with cost overruns in large infrastructure investments and partly with behavioral issues concerning investments in real estate. The introduction will therefore mainly deal with cost overruns and in a minor way deal with behavioral issues.

1.1 Cost overruns

There are three main questions in the research field that is covered by this study and they are ‐ Are there cost overruns? ‐ What explains cost overruns? ‐ What can be done to prevent cost overruns? In this thesis five articles have been written that are related to these main questions. In one paper (paper 2 in the list below) it is shown that it can be rational for construction client to design procurement in such a way that cost overruns can be expected. In another paper (nr 3 in the list below) a method used to prevent cost overruns are evaluated. In the third paper on cost overruns (nr 4 in the list below) a new framework for explaining cost overruns is presented. The final paper on cost overruns (paper 5) focus on policies that can prevent cost overruns. These papers were written within a research project on cost overruns in infrastructure projects financed by Formas.

1.2 Behavioural economics

The first article is based on my licentiate thesis and deals with Behavioral Economics. The background was my own experience as a tenant representative for the telecom company Ericsson during the years 2000‐2001. At the end of 1990‐s there was high demand from tenants to rent office space in Kista. The rents went up and the vacancy was low. In this situation a substantial amount of production of office space started. Among those projects were Kista Science Tower and Kista Entre and altogether the production was approximately 250 000 square meters, which was 25% of the total amount of office space in Kista. To construct a new office building takes approximately three years and before this new office space were ready for the market there had been an IT crash in 2001. With both a substantial increase in supply and a decrease in demand the rents fell and vacancies increased. In the light of these problems many questions arose: Was the decision from clients to build in 1999 rational? If not, were they irrational and in what sense? These are the questions that article 1 are dealing with.

1.3 Organization of thesis introduction

This introduction is organized as follows. In section 2 a selection of earlier research on cost overruns will be presented. The literature is presented in relation to the three main questions; are there cost overruns, what explains cost overruns and what can be done to prevent cost overruns. As the thesis can be said to deal with decision making in general, section 3 presents two theoretical frameworks that has influenced the work. The first is standard neo classic economic theory and the second is behavioral economics. The first framework deals more with how humans and companies should make decisions in order to reach their goals while the second deals more with how people actually make decisions.

(10)

In section 4 my general view of a scientific research process is presented. It covers research questions, methods, results, limitations and finally what can be seen as a contribution. This framework is then used in section 5 to present the articles that are included in the thesis. Finally in chapter six some general conclusions are made from the thesis.

2. Cost overruns from a paradigm perspective

In this section an attempt to describe some of the research that has been done within the field and also to try to say something about what could be seen as scientific proofs. According to Thomas Kuhn, see Chalmers (1995), science develop through different steps, pre‐science, normal science and revolutionary periods where science change from one paradigm to another. I regard the theories around cost overruns as pre‐science or maybe normal science with a couple of competing paradigms. I will describe these paradigms in this section and use them in section 6 when I discuss the contribution of the thesis. The paradigms will be formulated in accordance with Karl Popper´s view of science as statements that can be falsified at least in principle. I will not be as demanding as Popper, and disregard a hypothesis if an observation says something else instead I will rather treat it as an anomaly to the ruling paradigm.

2.1 Core idea 1 in the paradigm : There are systematic cost overruns in

infrastructure projects.

Maybe the most exhaustive work has been done by Flyvbjerg et al (2002) and Flyvbjerg et al (2003) were they surveyed 258 projects involving both rail and roads. They found that approximately 9 out of 10 projects had cost escalations and the likelihood of cost overruns was 86%. They also found that actual cost were on average 28% higher than forecasted costs. In the research field on this topic there are a lot of case studies that support Core idea 1. Here are just a few examples. An early study is Hall (1979) where he analyzed cost overruns in projects such as Opera House in Sydney. Following Hall a couple of case studies can be found in the research literature, for example Fouracre et al. (1990). They analyzed construction of metros in undeveloped countries and found that nearly all out of 13 cities experienced cost overruns. Morris (1990) looked at public sector projects in India, including coal‐, steel‐, agriculture projects etc., and also projects for constructing railways. He found that there were in general cost overruns for these projects . Morris points at a couple of factors for public sector overruns and the most important factors according to Morris were bad project preparations leading to scope changes. Nijkamp and Ubbles (1999) analyzed five projects in the Netherlands and three projects in Finland. Their conclusions were that in these projects there were cost overruns in all projects. Odeck (2004) investigated road projects in Norway constructed over the years 1992‐1995. He found cost overruns on average 7, 9% with a range from ‐59% to 183%. He found a correlation between small projects and cost overruns, the smaller projects the larger cost overrun. There were also a correlation between completion time and cost overruns and between regions and cost overruns. Looking at these and other studies there seem to be enough evidence to say that there are systematic cost overruns in general. To capture Poppers falsification the above papers do not try to falsify the paradigm, rather in an inductive way try to prove Core idea 1, which is perhaps not the

(11)

correct way to deal with a hypothesis or paradigm. On the other hand, the empirical studies could just as well have found that there were no cost overruns and then falsified the core idea.

2.2 Core idea 2 in the paradigm: Systematic cost overruns are due to

optimism bias and strategic misrepresentation.

The most convincing theory supporting this idea is Flyvbjerg et al. (2009) were they argue for two explanations which they call delusion, which I interpret as optimism bias and deception which is strategic misrepresentation. They argue that delusion is a cause of behavioral traps which they call the Planning Fallacy and Anchoring and Adjustment. Deception is caused by the principal‐agent relationship between the parties within organizations that are responsible for infrastructure projects. Flyvbjerg et al. (2010) still argue for strategic misrepresentation which they here call political explanations, and optimistic bias which they here call psychological explanations. In the paper they once again has principal agent theory and prospect theory as frameworks for explanations. In the paper they add two more explanations that is economical and technical explanations. Economical explanations, which they state in their paper, are very similar to strategic misrepresentation. Technical explanation is also mentioned in the paper as a cause of cost overruns. It is not included as a core idea here because it is more of a description of what happens – “there are technical problems” – but as argued in paper 4 it is difficult to see this as a cause. Core idea 2 is strengthened in Flyvberg et al. (2010) by their matching of causes to theories such as principal agent theory and prospect theory. Support for Core idea 2 can further be found in an article from Wachs (1989) who discuss cost overruns due to planners lying (strategic misrepresentation) in order to get their project approved. He gives personal experience in the matter and even if this cannot be seen as scientific evidence it is still an observation that is consistent with the theory. The case study in Fouracre (1990) contribute to Core idea 2 by pointing out at a range of explanations such as important changes in project, too optimistic cost estimations and bad management of cost control. Kain (1990) also presents results in line with the Core idea, as he gives a picture of planners trying to get a railway project in Dallas approved by giving false estimations on travel forecasts and cost estimations. Flyvbjerg (2007) give support for the strategic misrepresentation were he states that planners and promoters deliberately misrepresent costs, and also benefits and risks, in order to increase the likelihood of getting the projects built. There are no paper challenging the first core idea that there are Systematically cost overruns, but there are several papers who discuss why there has been cost overruns and that focus on other factors than Core idea 2. For instance Nijkamp and Ubbles (1999) who perform case study in the Netherlands and Finland and found cost overruns depending on changes in project, changes in price and bad estimations. Other more recent examples can be found in paper 5. In summary Core idea 2 is somewhat loose but still the best explanation to systematic cost overruns presented so far.

2.3 Policy implication to Core ideas in the paradigms

The paradigm states there are systematic cost overruns and that they depend on strategic misrepresentation and optimistic bias. What are the policy implications if one wants to prevent cost overruns?

(12)

Flyvbjerg et al (2009) give several recommendations to prevent strategic misrepresentation depending on which principle‐agent relationship that is at stake. In an infrastructure project there are several principal‐agent relationships. One relationship is between the taxpayers as a principal and institutions proposing projects as agents. One possible policy is then that those proposing the project should share the financial responsibility. If the focus is on the relationship between the government as principal and planners and bidders as agents, the solution could be rewards related to outcome and more critical evaluations, but also financial participation by the bidder/contractor by using special contracts. There could also be a public review of construction costs before making decisions. Another policy recommendation preventing strategic misrepresentation and optimistic bias is by using Reference class forecasting. By using outside views to a higher degree and comparisons with final costs in other similar projects the risk of cost overruns can be reduced. A third way of preventing cost overruns (but not systematic cost overruns) is by using better tools for analyzing risks and technical challenges. Examples are systematic work with planning the project with systems such as the Successive Principle. Paper 5 discusses policies more in detail.

(13)

3. Theoretical frameworks

Section 2 has briefly explained the research area concerning cost overruns. In this section I will present the theories that are used in the thesis as a framework for the analysis. There are, mainly, two theories that has been used: Neo‐classical theory which can be seen as a more normative theory about how rational people should make decisions and behavioral economics which focus on how people actually make decisions. Both frameworks are used in all articles. I will first present Neo‐ classical theory and then behavioral economics.

3.1 Neo‐classical theory

The base of neo‐classical economic theory is, when resources are limited, to answer the questions: What will be produced? How much will be produced? How will it be produced? and For Whom will it be produced? To answer these questions the theory is based on a number of assumptions about the consumers and the producers. The presentation of Neo‐classical theory is short as this theory can be assumed to be well‐known. 3.1.1 The assumptions about consumers/households/individual actors The individuals are supposed to always want to maximize their utility, which can be interpreted as their happiness. Other assumptions concerns their taste and it is assumed that they are complete, that is they always know if they like, dislike or are indifferent between different goods and combinations of goods. They are also non‐saturated, that is more of a good is always better than less. A third assumption is that the taste must be transitive, that is if a person like good X better that good Y and good Y better than good Z, then the person must also prefer good X before good Z. A fourth assumption is the assumption of continuous substitution, that is if a person has two goods and lower the amount of one of the goods a higher amount of the other good will always be able to compensate the person so that the same utility is reached once again. A fifth assumption is the assumption of decreasing marginal substitution. This means that even though it is possible to compensate for a loss of one more good, the less that is Ieft of that good, the more of the other good is needed to compensate for the loss in order to maintain at the same utility. A final assumption is that the consumers have a limited amount of money to spend, there is a budget restriction. In the thesis this theory can be found in many of the assumptions about how actors will react, for example if incentives are changed for people involved in the management of a project. 3.1.2 The assumptions about producers Just as in the consumer theory above, the core idea in production theory is that firms are rational and try to maximize profits. In order to maximize profits and with the assumption of a decided, specific, produced amount and fixed price of the good to a customer. It could be for instance the construction of a bridge for a certain price. Then the goal for profit maximization becomes to minimize costs, that is min , min, such that ,

(14)

Here is the total cost, and are the amount of input good one and two and the price for input goods are and . We simplify and assume that only two input goods are needed to produce the final good in amounts. The solution to the minimization problem is that the amount of inputs that the firms’ will use depends on input prices and technology. This can be viewed in the diagram below were isocosts presents total costs at different levels of quantity of input goods and isoquants presents the amount of the final good. In the diagram below the amount q can represent a bridge.

The optimal solution for the firms in the above example is to use, ∗ of input good one and ∗of input good two. This optimal solution will change if there are changes in ‐ Input prices ‐ Amount of the final good ‐ The technology In the thesis this theory is especially used in paper two about how cost overruns can be a cost minimization solution.

3.2 Behavioural Economics: A general introduction

I will mainly in this short presentation use the work of Kahneman and Tversky and as a source for the non‐technical description use Kahnemans book “Think fast and slow” (Kahneman 2011)‐ Behavioral economics is about how individuals make decisions. In this section I will first present the simplified theories of how the brain works. Then there is a presentation of the most prevalent decision mistakes that human make. And finally a short criticism of economics from a behavioral economics point of view. In the thesis the behavioural theories will primarily be used in article 1 about determinants of investment behavior and in paper 4 about causes of cost overruns. Figure 2. Cost minimization depends on input prices, amount of final good and technology. Quantity Isoquant ∗ Isocosts Quantity ∗

(15)

3.2.1 System one and System two The brain can be summarized as working with two systems, or two processes. System one is the automatic thinking that is always turned on. It cannot be turned off. It demands no effort and can do simpler activities such as calculate 2+2, end the sentence “Bread and …”, drive on an empty road etc. System two is the controlled thinking which we often use to solve problems. Among its job assignments are to control system one. And it can do controlled activities, which system one is not able to do. System two is lazy and demands effort to function. With this effort it can do advanced calculations, pay attention to someone talking, be aware of its own behavior and do tricky parking with a car. The problem for humans it that both system one and system two are limited when it comes to decision making. One of the problems is that humans sometimes are relying on intuition which Kahneman (2011) state as recognition and as a process of system one. There are many shortcomings in human decision making, below some of the most important will be presented. 3.2.2 Problem with decision making The problem with our decision making can be summarized in two steps: 1. We have trouble dealing with certain types of decisions and therefore make bad judgments. 2. We are overconfident about our judgments. These two statements mean that we are very confident about our bad decisions. In this short presentation I will list a couple of bad judgments and I will also describe the problem of overconfidence. The problem with small sample The problem with small sample is that humans have a tendency to draw conclusions from too small samples. The problem is that a small sample has a tendency to construct extreme values more often than large samples. A large sample will more resemble the population but a small sample may not. If we are not aware of this we might draw too far reaching conclusions. If two groups are drawn with four in each and one group has four successes and the other group two successes and two failures, we cannot draw conclusions that the first group is more successful, as it is a high probability that it is chance that is making this difference. But humans are too quick to draw conclusions. Kahneman explains this phenomenon with system one characteristics, the Halo effect. The Halo effect is a system one tendency to think that because one thing is good with a subject, everything else with the subject must also be good. If you think a person is a good speaker the Halo effect makes people think that person must also be a nice person, intelligent etc. It is the same effect as when looking at a small sample. The problem with anchoring The problem with anchoring is that humans have a tendency to be effected in their judgments from what are actually not logical factors. The classical example performed by Kahneman and Tversky (1973) was to ask two groups of students how large part of the members in UN came from Africa. Before they answered they saw a roulette, which was prepared by Kahneman and Tversky, only to land on number 10 and 65. The group who saw the number of 10 on average answered 25% and those who saw the number 65 on average answered 45%. This example has been tested a lot, for

(16)

example on how real estate brokers are affected by anchoring when they are to estimate market values (see below). According to Kahneman the problem with anchoring can be described with both system one and system two. With system one it is related to the primacy effect, that humans are subconsciously affected without knowing it. System two explanations is that humans try to compensate for the anchoring effect but has difficult to do it enough. Heuristics The problem with heuristics is that when trying to answer a question (the goal question) humans have a tendency to instead answer another question (heuristic question). For the question “How pleased are you with your life” (goal question) are substituted for “How happy am I right now?”. Heuristics can also be 3‐dimensional, se figure 3. This is an example of how difficult it is to avoid the heuristic issue. “Which one of the three persons in this two dimensional drawing that is the largest” is the goal question”, but the heuristic question becomes: “ This is a three dimensional picture, which one is the largest.” Your answer in the heuristic question will be the person to the right. The correct answer is that they all have the same length. Another example is the availability heuristics, when the judgment of frequencies is made with how easy it is to remember examples. The goal question might be: “How common is divorces among celebrities” is replaced with the question: “How many and how easy can I remember actual divorces among celebrities”. One explanation to the heuristics is according to Kahneman (2011) two characteristic that system one has, the mental shotgun and the matching of strength. The mental shotgun is the characteristic that system one always has to calculate; as soon as we wake up there are automatic calculations and system one create the three dimensional space we see, the form of different subjects, their place in the room etc. When system two is asked to answer a controlled question it is difficult to restrain system one which delivers a surplus of calculations, called by Kahneman the Mental Shotgun. The matching of strength is the ability of system one to compare different appearances. If murder was a color it would be darker than the burglar etc. Another example compares the sentence: “Julie could read when she was four years old”. It is not difficult for us to answer the following questions: “What kind of income does Julies reading ability correspond to?” or “How tall a man that has equal height as when Julie is mature?” System one is able to match these different aspects. Figure 3. Which one is the largest, source Kahneman (2011)

(17)

When a difficult question is asked (the goal question) system one start delivering many different answers (mental Shotgun) and the answers are compared to the original question and even though they do not answer the original question system one finds an answer that is nearly similar but actually answers another question, the heuristic question. Overconfidence Above some of the problems with decision making are described. The problem gets worse because humans think they make good decisions, they are overconfident. The early example of overconfidence was Albert and Raiffa (1982). A description of it is made by Taleb (2010) were the persons are asked to judge something. It could be size of a city, the number of books sold each year etc. The persons were asked to give an interval where they were sure that the correct answer with 98% would be. “I am sure that with 98% this city has between 100 000 and 150 000 inhabitants”. It was found that 45% of the persons missed the true value with their interval instead of expected 2%. This experiment has been tried on different categories of people and the common conclusion is that people are generally overconfident about their judgments. The phenomena of overconfidence are to some extent explained by Kahneman (2011) with the illusion of understanding which is based on the system one characteristics of building causality, the what you see is all there is (WYASITI) and the Halo effect. 3.2.3 Behavioral Economics and Neoclassical Economics This part of the text will present the critical view it has on the assumptions of rational behavior within the field of Economics. In economics it is stated that in an uncertain world individuals will make decisions in line with expected utility, which is weighing together the final outcome with the probability: is calculated and the alternative with the highest expected utility ought to be chosen by a rational individual. The problem with this is both the probability figure and the stated utility. Prospect Theory The way the concept of utility is used within economics does not take into consideration the fact that people use reference points. In economics it is assumed that utility only depends on the absolute level of consumption, which would imply two persons with the same preferences and the same amount of money, e.g. 5 million, would have the same utility. But this is not the case. Assume one person the day before had one million and the other nine million and today they have five million. According to Kahneman & Tversky, they have different reference points and will of course not have the same utility, even though they have the same amount of money. The reference point is an important aspects lacking in economics. Prospect Theory also deals with the observation that the pain of a loss is higher than the joy of a gain, and that behavior is asymmetric between losses and gains. Decision Weights The problem with probability is that people do not behave as linear as one might suspect when it comes to probabilities. This is illustrated in table 1 below. On the upper row the probability of different outcomes are stated, and on the lower row the decision weight for humans at different probabilities are presented, in line with experimental evidence. This is an example, but describes a situation where an actual probability of 1% has a 5,5% decision weight for individuals, and at the

(18)

other end a 99% probability has only a 91,2% decision weight. Humans have a tendency to overweight low probabilities and underweight high probabilities. Prob. 0 1 2 5 10 20 50 80 90 95 98 99 100 Dec. 0 5,5 8,1 13,2 18,6 26,1 42,1 60,1 71,2 79,3 87,1 91,2 100 The explanation according to Kahneman (2011) is, for low probabilities, the possibility effect and for high probabilities, the certainty effect. The possibility effect is when you go from a zero chance to at least a chance at all, even though it is very low. This is why people buy lottery tickets. Going from zero chance and buying no ticket to a, at least, possibility to win is a gigantic step. The certainty effect is going from a 99% chance to 100%. That is going from there is a risk of failure to no risk of failure at all. This is something that people value more than one percent change in decision weight.

3.3 Behavioural Economics in the real estate sector

In this section I will shortly describe the research that has been done on the Real Estate Economic field based on behavioral Economics, using primarily Diaz III and Hansz (2007), Salzman and Zwinkels (2013) and the literature review in Diaz III (1999). The research is not very large and most of the research has dealt with anchoring, as described above. Diaz III and Hansz (2010) found that anchoring, they call it reference points, affected the property value estimation of the participants in the experiment. Gallimore (2007) also found the effect of anchoring among real estate valuers. Black (1997) found anchoring effect on negotiators, who tend to anchor on the asking price for the real estate. Other behavioural issues has been mentioned and tested for example availability heuristics (the tendency to judge frequencies and probabilities with how many examples that come to mind and how easy those examples were remembered) and representativeness in Osmond et al. (2013). They could however not fully confirm those biases. Gallimore (2007) deals with recency effect but cannot prove that the biases exist. Gallimore and Grey (2011) opens up the interesting field of sentiment, or intuition. They find that intuition has a large role in many real estate investments. The interesting thing with intuition is that Kahneman (2011) states that many times humans uses intuition but that it leads to bad decisions. Here further research could be interesting. I restrict myself her from dealing with Client Pressure and Appraisal Smoothing as I do not consider this a behavioural economic issue, even if can be seen as leading to “wrong” decisions. Behavioural economics deal with cases when the agents do not act in a rational way. Client Pressure deals with the pressure that the client to a real estate valuer put on the valuer, which might lead to higher market values than otherwise. I consider the real estate valuer as acting rational even though the pressure from the client influences the value. Also Appraisal smoothing is not a behavioral issue. Smoothing refers to a situation where real estate valuers have a tendency to smooth out estimations of market values. This is due to the time between comparable sales and actual date for market value estimation. The valuer has to seek comparables in the past to lower the noise in the valuation. This is also rational even though the market value deviates from the theoretically correct value. Let me finally also state that game theory is also not Table 1. Above the probability and below the decision weight it is given.

(19)

behavioral economics. The famous prisoner’s dilemma makes the participants worse off than if they could communicate, which they cannot, and in that circumstance they make the next best solution, but given the structure of the game they act in a way that is rational.

(20)

4. Framework for presentation of the articles

In the next section each article will be presented. The presentation follows the same presentation format that will be presented in this section. The different steps are the Research question, the Method, the Results, the Uncertainty and the Contribution, see figure 4. This might seem like a linear process and even if this is not the case in practice, it still seems to be the best way to present the research reported in each article. This approach is influenced by Teorell et al (2007) and a more extended version of research step is presented in Brunes 2014 and writing that was also a way for me to upgrade my own knowledge of research and problems in different stages in the research process.

4.1 Research questions

The first task in research is to find an interesting research question. It an applied research field like real estate economics it should be interesting for both the scientific world and for the practitioners. Research questions can be divided into three major groups, describing, explaining and normative questions, see table 2. Describing What? How? Explaining Why? Valuing How should it be? In a descriptive study the researcher tries to answer questions like: How big is the stock of real estates in Stockholm? How many lives in condominiums? How much did a certain project cost? Explanatory questions focus on why a certain state has occurred. Why is the stock of real estates in Stockholm a certain amount? Why do some people live in condominiums? Valuing questions focus on how it should be, How many buildings should there be in Stockholm? How many people ought to live in condominiums? Especially in engineering research questions about how something should be done to reach a certain goal is important. The chosen type of research questions decides what kind of arguments that is needed and what kind of method that might be the most preferable. Research question The Results The Uncertainty The Contribution The Method Figure 4. The steps in research. Table 2. Type of research questions.

(21)

4.2 The Arguments

To be able to answer the research questions a number of conditions has to be fulfilled. Here three conditions will be discussed. The first condition is that correct concepts and measurements are used. The second condition concerns arguments for causality. The third condition is related to the representatives of the population. Condition 1 – Concepts and measuring The first conditions to construct definitions for different concepts and to find a way to measure them. When definitions and measuring units has been decided, the measurement must be correctly performed, see table 3. Here three instruments are used to measure temperature for boiling water at four different occasions. Instrument 1 has high stochastic variation that is the measure varies over and under the correct value. It has high volatility. Instrument 2 has low volatility but it deviates from the true value in the same direction. Instrument 3 has high validity and reliability, it measures the water temperature correctly each time.

Instrument Temp Temp Temp Temp Average Validity Reliability

Actual temp. 100 100 100 100 100 ‐ ‐ Instrument 1 80 120 90 110 100 Hög Låg Instrument 2 81 79 80 80 80 Låg Hög Instrument 3 100 100 100 100 100 Hög Hög Condition 2 – Causality Causality can never be proved but there are arguments that combined can point at a causal relation between two phenomena, let us call them X and Y. These arguments are contrafactual difference, isolation, time, and mechanism of causality. Contrafactual difference states that if X happens than Y must also happen and if X does not happen then Y should not happen. Isolation state that X is causing Y and nothing else, see figure 5. Here it is not X that causes Y but rather Z that causes both X and Y. There is a spurious (false) causality between X and Y. It seems that there is causality but it is not. Table 3. Measuring correct or not. Figure 5. Spurious effects when isolation cannot be shown. X Z Y

(22)

The third statement is that X must in time precede Y, see figure 6. Here changes in X are preceding changes in Y. The fourth condition for arguing for causality between X and Y is to find mechanisms between X and Y, see figure 7. Here this is shown by arguing that between X and Y events W and V connect X and Y. As was stated in the beginning, causality cannot be proven but if arguments of contrafactual difference, isolation, time and mechanisms can be shown, at least there are good arguments for believing that there is a causal relation. Demand 3 – Empirical generalization Often a whole population cannot be measured as above, but a sample has to be chosen, see figure 8. Here the samples of observations, the blue dots, are used to draw conclusions about the whole population. To be able to perform generalization the sample must have high internal validity, which Figure 6 Order of time. X / Y Time X _Y Figure 7. Mechanism connecting two events. X Y X W V Y Figure 8. A sample of buildings to describe the whole population.

(23)

is fulfilling the demands of causality described above, and external validity that is no systematic or stochastic sample bias.

4.3 The methods

Research methods in economics can be divided in to two main groups, methods dealing with drawing conclusions from data (induction), and methods dealing with reasoning (deduction), see figure 9. Below there is a very short presentation of the two broad fields. Induction Induction uses observations to draw conclusions. There is in general three methods to collect data, that is historical data (sales of houses, rental agreements etc.), surveys (asking people about a subject) and observation, when you observe how people and companies behave on the market. Various more or less advanced statistical methods are typically used to identify patterns in the observations. More general conditions for drawing causal conclusions are described above. Deduction Deduction use assumptions and rules to draw conclusions see figure 10. The method uses premises and logical rules to draw conclusions. Premises can be divided into theory (laws, rules) and initial conditions which are used to perform explanations or forecast. Theoretical models in economic can be used as an example. In the models various assumptions are made and then results are deduced using mathematics and logic. Theory Observations Induction _Deduction Figure 9. Induction and deduction. Theory Initial conditions Forecast

All individuals are rational.

House A is identical to House B but cheaper

Individuals will buy house A Figure 10. Left is the structure of deduction and to the right an example.

(24)

4.4 The Results and Uncertainty

The results should answer the research question. How well the research question has been answered depend mainly on how well arguments for the results can be stated. This is directly related to uncertainty, the less arguments/evidence for a result the more uncertainty. This is illustrated in figure 10. The strength of the results depend on how well has concepts and measurements been stated, (construct validity), how well has the evidence for causality been stated (intern validity) and how well has the generalization from a sample been motivated (empirical generalization).

3.5 The Contribution

Finally if the results are good and uncertainty low the results might contribute to develop science within the specific field. On a more general level the question is how is science actually developed? According to Chalmers (1995) there are several different approaches in the philosophy of science. The first is the inductive approach with induction and deduction presented in section 3.4. That approach is sometimes criticized for being naïve in the use of observations, and the main argument against the inductive approach is that observations are always uncertain. Another argument against is the problem with inductions – with many observations there is still uncertainty about the conclusion. There can still be an observation in the future that says the opposite to what has been found so far. The second approach is the falsification approach were a theory is viewed as a temporally truth and should be stated in a way that it can be tested and thereby falsified. “The earth is round” is a theoretical statement in line with the falsification approach. If the theory is falsified then a search for a new theory must be started. Or maybe there is a new theory developed from the falsifying evidence. The problem with falsification is, historically, that there has always been some observations that was not consistent with a certain theory and if the researchers had used a crude version of the falsification approach the theories would just have been rejected and not developed further. Figur 11. The uncertainty of the result. Construct Validity The strength of the result Internal Validity Empirical generalization

(25)

The third approach, according to Chalmers (1995/2013), is to see research as a program, an approach initially stated by Kuhn and Lakatos. They have different view on how to define science but a similar way to approach the subject. I will below follow Kuhn´s theory of paradigms, see figure 11. To the right is the theory with a core (the black area) of the paradigm The core is what the scientist assume is correct about the topic, and there is no attempts to falsify the core. It is surrounded by a white area where an uncertainty about the theory is located, and where research results which deviate from the core has been placed for further treatment. The scientist working within the paradigm tries to find explanations for these deviations that is consistent with core of the theory. Economists might for example try to show that what looks like irrational behavior might actually be rational given the information and expectations of the actor. However, if the deviations become many there might be a change of the paradigm and the black core is changed, there is a paradigm shift and a scientific revolution. The change from neo‐classical to behavioral economics can be seen as such a paradigm shift. To the left are research done within the field, the plus sign means that the results of the study is consistent with the black core and makes it even stronger, it improves the present paradigm. A minus means that the research deviates from the present paradigm. Research is a continuing activity with results questioning and improving paradigms. The contribution for the articles in the essay will primarily be compared to the achieved contribution within the field for the questions: Are there cost overruns? What explains cost overruns? What can be done to prevent cost overruns? They will also be related to the controversy between standard assumptions of rationality and the behavioral theories presented above.

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + ‐ ‐ + + + + + + + + + + + + + + + + + + + + + + + + + + Figure 11. To the left are many research results and to the right is the theory within the field.

(26)

5. Summary of Articles

In this section I will make a short presentation of each article in the essay using the format I introduced in previous section.

Article 1 – Overbuilding in office markets: Are behavioural aspects

important?

The aim The aim of this paper is to find out if investors within the Swedish office market have been influenced by myopic behavior, herd behavior and overconfidence. Myopic behavior means that decision makers only look at what is happening on the market at the present time, and what the present trends are. The decision maker has then extrapolated this trend and assumed that it will conintue into the future. These forecasts are done even if you in an historical view can see that trends typically are short, see figure 11. Herd behavior means that the investor follows other investors even though their own indications say that they should not. Finally overconfidence is when investors systematically overestimates the expected performance of their projects. The arguments / The method The hypothesis that decision makers seems to be myopic is the article first tested by looking at if the decision makers are too focused on present market value instead trying to figure out the future market value when the construction is completed. If it takes three years to complete an office building and there is a connection between the completion of office buildings and market value three years earlier it can indicate that the decision makers were myopic: they were looking at the current market value instead of trying to predict the future market value. In order to evaluate the rationality of the investor construction costs have to be taken into account. An investor will not start construction if the construction costs are higher than the market value. To capture the relationship between market value and construction cost Tobin´s Q (TQ) was used. This is Figure 11. Myopic view means that the investor thinks the present trend will last long into the future. Time Office rents Today

(27)

defined as the quota between market values and construction costs, see the formula below. If TQ is higher than one there are incentives to construct.

To find out the correlation between TQ and the construction of office buildings (P) a regression analysis was performed with the formula

P

_t

  

 

TQ

_{t m}





_t. The equation was estimated with different time lag is in order to see for which time lag Tobin´s Q performed the best. The hypotheses about the different behavioural factors were also tested by using a questionnaire and three case studies. The results The estimation showed that the highest correlation were between the volume of construction at time t and TQ three years before completion and increase in the supply (P), see table 3. Parameter t

TQ

_t_₁

TQ

_t_₂

TQ

_t_₃

TQ

_t_₄ Constant coefficient 22354 1613 14175 -18567 -16258 TQ coefficient 7362 34549 14346 59101 55011 t-test TQ 0,29 1,42 0,61 2,48 2,21 Korrigerat R-testl -0,04 0,05 -0,03 0,21 0,18

Autocorrelation in the residuals No Yes No decision No No decision

There seems to be contrafactual causality between TQ at t‐3 and P at time t. Also the time factor is correct where a high P is preceded by a high TQ. This can point at causality that TQ at t‐3 explains P at t which would indicate myopic behavior. It was however difficult to clearly find evidence for herd behavior and overconfidence in the case studies and the questionnaire. Both however supported the hypothesis of myopic behavior. Uncertainty First there are uncertainties with measuring both market value and construction cost. Secondly it can of course be discussed if the relationship between TQ and completed construction is a good tool to capture myopic behavior? Maybe TQ catch other aspects, maybe the investors were not influenced by myopic behavior and still made the decisions? Table 3. Regression results with different time lags.

(28)

Secondly, there is uncertainty in the causality between TQ and P. It was not stated that there is isolation between TQ and P. Was it really the TQ that influenced P? Maybe it was some other factor influencing. There might be a substantial option value to construct even if there will be large vacancies initially. Also, just looking at the relation between TQ and P will not show any mechanisms and a chain of causality from myopic behavior to completion of office buildings. Third, there are uncertainties with the statistical sample. The sample was not stochastic but strategic, starting from the available transactions on the market. There might be problems with bias in the sample. It is problematic to generalize the results from this sample to the whole population of decision makers. Fourth, there is an uncertainty concerning measuring in the case studies and questionnairs. How certain can the respondents be that it was really any of the hypothesis that influenced the decision makers. Do the respondents understand the questions and is it correct to interpret their answers in the way that was done. Contribution The contribution of the study can be discussed on to levels. First, if we look at the field of real estate economics research as it was in 2005, behavioural aspects were hardly discussed at all and this was one of the first attempts to test some of the behavioural hypothesis empirically. There was only a small literature in real estate economics concerning behavioral explanations at that time. Secondly, even if there is a lot of uncertainty and that more far reaching empirical conclusions should not be drawn, still the results give support especially for the hypothesis of myopic behaviour. Going back to the figure describing paradigms in figure 11, the study would be a plus sign in the relation to at least one of the core hypothesis in the behavioural paradigm, in relation to a theory that behavioural issues are important for real estate investment decisions. bute with behavioral issues as explanations to real estate investment decisions.

Article 2 – Quantity Choice in Unit Price Contracts Procurements

Aim The aim of this article is to find out if it can be rational for a client to design procurement in such a way that it expects cost overruns when using a Unit Price Contract in a Design‐Bid‐Build method for project delivery. Argument/Method Our analysis starts with a model, see figure 12, with a client who procure according to the Design‐ Bid‐Build method. This means that the client before procurement decides what amount of q, for instance mass transported, that should be included in the contract with, and how much effort, , the client should put to monitor the actual amount of q that the contractor carries out.

Every unit, q, will after the procurement get a price of p sek/unit. The client doesn’t know the actual amount that the project needs. They make a judgment that it is between and and when construction has started the actual quantity, will be revealed to the contractor.

(29)

If the client procure the quantity, q, which exceeds actual outcome, , the total cost will be 1 were C is the cost of monitoring and where α is between zero and one, where zero implies no monitoring of the contractor and one means full monitoring. If there is no monitoring the cost will be , as the client pays according to the contract even if the actual quantity is lower. If the client underestimate the amount of material, that is , it is assumed that there is a renegotiation with the contractor. Then a renegotiation cost, R, and a new price, (p+γ), for the remaining amount ( ‐q) and the final cost will be .

Figure 13 shows how the cost vary depending on whether actual volume, , exceeds or fall short of ordered quantity, q, and if no monitoring is performed, α=0. At the final amount fall short of ordered, q, which implies a total cost of . At the actual amount exceeds the ordered amount and there will be a renegotiation, O, and a new price that will lead to cost .

The question is where on the interval, and the client would set q in the contract? The answer is simple, where the expected cost is lowest. An important question is then at what quantity, q, the cost will be lowest? That depends on the probability that renegotiation occurs. Let’s start with the case were there is no monitoring and then later with the case of monitoring. The clients optimizing of q when there is no monitoring. If there is no monitoring the optimal q will change according to table 4. Renegotiation Actual amount (Q) Cost

Figur 13. The total cost varies depends on q, and and α.

q

R q≥Q Figur 12. Time axis for the model. Procurement gives price p. Finished: Cost 1 Decide q and Construction reveals Q q<Q Finished: Cost

(30)

. Scenario *

q

1. Cost for renegotiation (O) and/or renegotiatid price (p+γ) increase. Increase 2. Expected price (p) increase. Decrease 3. Increased uncertainty, i.e. the

distance and increase

Uncertain

Clients optimizing with monitoring, α>0

If monitoring can be varied the optimal q will also vary, that is . With a high amount of monitoring the ordered amount can be, all other things equal, higher and vice versa. The client will then see when the contractor has finished and will know the actual amount. At the same time is monitoring a function of ordered quantity, . If q is low/high the constructor will put less/more resources on monitoring.

This is a dynamic system of equations which is analytical difficult to solve but easier to analyze numerically and the result from the numerical analysis is presented below.

The results – when will cost overruns be present?

If cost overruns are defined as the difference between final cost and contract cost, then cost overruns should be expected in this model. It is rational and optimal for the client to sometimes to have cost overruns in this case.

Cost overruns will occur in larger amount if optimal ordered quantity

q

* is very low. Because of that it is interesting to see how the optimal ordered quantity will change when there is a change in original price, renegotiation cost, renegotiated price or uncertainty interval, see table 5. Scenario *

q

Risk for cost overruns 1. Cost for renegotiation (O) and/or renegotiatid price (p+γ) increase. Increase Decrease

2. Cost for monitoring increase. Decrease Increase 3. Expected price (p) increase. Decrease Increase 4. Uncertainty increase, i.e. the

distance and increase

Uncertain. Maybe increasing Uncertain Cost for renegotiation and/or renegotiated price decrease: This creates an incentive for the client to order lower amounts and risk the cost of renegotiation and new price. This will lead to an increased risk of cost overruns. Table 4. Changes in q when other parameters are changed Table 5. Possible outcomes of q and cost overruns when four parameters are changed.