M4A2: Expected Value and Consumer Choices

(All Articles are attached. They must be used and TurnItIn is used to check all papers.)

Consumers’ choices are prey to subtle discrepancies that arise in cognitive accounting. Learning how and when you are prey to these discrepancies is an important step in improving your decision making.

As the readings for this module demonstrate, people value gains and losses differently under different scenarios. For example, contestants in a game show might choose a guaranteed $10 prize over a 50 percent chance of winning $20 despite the fact that the expected values are the same.

Using the readings for this module, the Argosy University online library resources, and the Internet, address the following:

What is mental accounting and how does it impact consumer decision making?
How might a company take advantage of consumers’ mental accounting? Give examples.
As a marketer, how might you frame certain decisions to benefit from the disparities that arise in one’s cognitive accounting?
As a consumer, how would you avoid the pitfalls posed by the inequalities of one’s cognitive accounting?

Write a 3–5-page paper in Word format. Apply APA standards to citation of sources. Use the following file naming convention: LastnameFirstInitial_M4_A2 ..

Assignment 2 Grading Criteria	Maximum Points
Examined the concept of mental accounting and how it impacts consumer decision making.
Adequately described how a company can take advantage of consumers’ mental accounting, providing appropriate examples.
As a marketer, explained how certain decisions might be framed to benefit from the disparities that arise in one’s cognitive accounting. As a consumer, explained how to avoid the pitfalls posed by the inequalities of one’s cognitive accounting.
Wrote in a clear, concise, and organized manner; demonstrated ethical scholarship in accurate representation and attribution of sources; displayed accurate spelling, grammar, and punctuation.
Total:	100

“Social loafing” describes the fact that people tend to work harder as individuals than as members of groups. The net result of this effect can be that groups are less effective unless incentives are in place to motivate individuals.

Another group pitfall is “conformity.” Conformity, as the word suggests, refers to matching one’s belief systems, attitudes, appearance, and behaviors to the perceptual norm of the group. This can have disastrous consequences for organizations when the majority view leads to poor decisions. Being complicit with the group point of view clearly puts limits on both creativity and the available frames.

Finally, similar to conformity, “groupthink” diminishes the values of individual creativity, uniqueness, and independent thinking. It occurs when insulated groups succumb to local loyalties and the related pressures to conform to the group. These conformance pressures tend to limit critical evaluation of alternative solutions in favor of consensus and conflict avoidance. Groupthink can lead to decaying judgment and efficiency. However, it is also true that groupthink, when appropriately facilitated, can accelerate the decision-making process and improve its precision.

Of course, groups are fertile incubators of creativity and thereby, new frames and previously unseen options. Groups benefit enormously from repeatable frameworks and centralized, policy-based rules. These tools streamline, or help manage, complex decisions as well as help alleviate the social pitfalls outlined. They also tailor decision-making processes to coincide with the strategies, structures, and even the incentive schemes of organizations so that managerial choices reinforce organizational strategy.

In light of the limitations of group decision making, it is important to understand where crowd-based solutions lose their luster, or even completely fail, and how and where decentralized decision making offers superior outcomes.

Roughly two-thirds of the US economy is attributed to consumer spending or roughly 10 trillion dollars in 2010 (US Department of Commerce: Bureau of Economic Analysis, 2011). This enormous amount of capital itself as well as how it is utilized is of constant interest to marketers, economists, and political leaders. Beyond the influences of social heuristics, how do your own internal metrics and accounting principles influence your decisions and how can those practices create opportunities? In other words, what is your mental accounting?

In 1980, Richard Thaler coined the term mental accounting in an attempt to describe how people categorize and quantify economic outcomes (Thaler, 1980). Eight years later, Shefrin and Thaler proposed that mental accounting is divided into discrete repositories; these “buckets” are current income, current wealth, or future income. Moreover these buckets have some interesting qualities from a cognitive and subsequently, accounting perspective (Shefrin & Thaler, 1988).

These accounts are largely non-fungible and the marginal propensity to consume from each account is different. The implications for this are profound; how utility is evaluated varies based on the mental account used. Likewise, perception of value changes at different points in time; people are prey to subjective frames.

Adding complexity to this mix, it is true that in your mental accounting applies two values to any transaction—acquisition value and transaction value. The acquisition value is the money you will trade to physically acquire a good or service while the transaction value is the price you place on getting a good deal.

Finally, the value placed on gains and losses differs between individual mental accounting. Similar to prospect theory, people have a tendency to skew utility in order to minimize losses and maximize gains.

Mental accounting of consumer-oriented decisions, coupled with the complexity of consumer choice (imagine all the different options and financing plans on the car purchase), and the departures taken from perfect rationality all influence consumer behavior. They determine when an individual chooses to act or postpone a purchase, how he or she perceives gains and losses, and how timing bears on the individual’s choices. Marketers especially want to leverage these predilections to frame one’s perceptions and choices in nonrational ways. However, one’s personal balance between self-control and buyer’s remorse is at stake.

Shefrin, H. H., & Thaler, R. H. (1988). The behavioral life-cycle hypothesis. Economic Inquiry, 26, 609–643.

Thaler, R. H. (1980). Towards a positive theory of consumer choice. Journal of Economic Behavior and Organization, 1, 39–60.

US Department of Commerce: Bureau of Economic Analysis. (2011). National economic accounts: National GDP. Retrieved from

http://www.bea.gov/national/#gdp

A Cost-Benefit Analysis of the New Orleans Flood Protection System

Stéphane Hallegatte1

Center for Environmental Sciences and Policy, Stanford University,

and

Centre International de Recherche sur l’Environnement et le Développement, Ecole Nationale des
Ponts-et-Chaussées

Abstract

In the early stages of rebuilding New Orleans, a decision has to be made on the
level of flood protection the city should implement. Such decisions are usually
based on cost-benefit analyses. But in such an analysis, the results are contingent on
a number of underlying assumptions and varying these assumptions can lead to
different recommendations. Indeed, though a standard first-order analysis rules out
category 5 hurricane protection, taking into account climate change and other
human-related disruptions of environment, second-order impacts of large-scale
disasters, possible changes in the discount rate, risk aversion and damage
heterogeneity may make such a hurricane protection a rational investment, even if
countervailing risks and moral hazard issues are included in the analysis. These
results stress the high sensitivity of the CBA recommendation to several uncertain
assumptions, highlight the importance of second-order costs and damage
heterogeneity in welfare losses, and show how climate change creates an additional
layer of uncertainty in infrastructure design that increases the probability of either
under-adaptation (and increased risk) or over-adaptation (and sunk costs).

Introduction

Six months after the deadly landfall of the category 4 Hurricane Katrina on New Orleans, there is
an active debate about the reconstruction of New Orleans and the design of its future flood
protection system (e.g., Schwartz, 2005; Bohannon and Enserink, 2005). Although the
reconstruction of New Orleans has been questioned by House Speaker Dennis Hastert and is still a
debated question (Hahn, 2005), this paper will assume that it will be eventually carried out and
focus on an adjacent question, namely the necessity of making the city flood protection system able
to cope with category 5 hurricanes.

1 corresponding author (hallegatte@centre-cired.fr). I am grateful to Philippe Ambrosi, Hans-Martin Füssel, François
Gusdorf, Minh Ha-Duong, Robert Hahn, Mike Jackson, Mike Mastrandrea and Jonathan Wiener for very useful
suggestions and advices on the form and content of this article. This research was supported by the European
Commission’s Project No. 12975 (NEST) “Extreme events: Causes and consequences (E2-C2)”.

The design of natural disaster protection systems is based on cost-benefit analyses (CBA; see for
instance Arrow et al., 1996), even though other decision-making frameworks have been proposed
(e.g., the precautionary principle). In a CBA framework, New Orleans would only benefit from a
flood protection system able to cope with category 5 hurricanes, compared to a system able to cope
with category 3 or 4 hurricanes, if the additional cost of the upgraded protection was lower than the
expected benefits from reduced flood damages. This is not certain to be the case, as surprising as
this might appear given Katrina’s devastating impact, and this article will show how two
evaluations can reach opposite conclusions.

To do so, we will first carry out a simple CBA using available data on the damages caused by
Katrina in New Orleans. As we will see, this first CBA clearly rules out any category 5 protection
system. Then, we will show how less optimistic assumptions on anthropogenic perturbations of
environment and considering additional processes, namely second-order impacts, discount rate
choice, countervailing risks and side-effects, risk aversion, and damage heterogeneity, can change
the terms of the analysis and potentially justify the implementation of a category 5 system.

A first Cost-Benefit Assessment

To carry out a CBA of a category 5 flood protection system in New Orleans, one needs to assess
the cost, C, of such a system, and its expected benefits, B. On the one hand, assessing the cost, C, of
an upgrade of the protection system is not that complicated, even though it requires a precise
definition of the system and an assessment of its construction and maintenance costs. In the very
early stages of rebuilding, state officials evaluate the cost of category 5 protection at about $32
billion, compared with a $1 billion cost to restore the initial category 3 protection (Schwartz, 2005).
Precise information on the cost of category 4 protection is unavailable but we will assume that it
would be about $5 billion, making the additional cost of a category 5 protection $27 billion.

On the other hand, assessing expected benefits is much more problematic, as one needs to take into
account benefits of various natures (avoidance of casualties, injuries, economic losses,
psychological trauma, etc…) impacting different groups of people and possibly lying far in the
future. This aggregation problem has been largely discussed; see, for instance, Adler (2005) or
Schneider et al. (2000) for aggregation issues between different categories of impacts and Portney
and Weyant (1999) for issues concerning inter-temporal aggregation.

Regardless of these important problems, benefits can be defined as the net present value of the
expected amount of damages avoided by the protection system upgrade. These benefits can,
therefore, be calculated as the discounted sum, for each year from now through the lifetime of the
protection system, of the annual probability that a category 5 hurricane hits New Orleans multiplied
by the difference between the damages of such a hurricane on a category 4 versus a category 5
protection system. This difference is discounted to take into account the fact that the same benefit is
valuated at a higher value when it occurs in the near future rather than further in the future.

The values of three parameters are thus necessary: the discount rate (δ), the probability of
occurrence (p), and the amount of avoidable damages in the year n (dn). From them, expected
benefits, B, are easy to calculate2:

(1)

Assuming that damages are growing over time at the same rate as economic growth, g, which is a
conservative hypothesis considering the current growth of economic losses due to natural disasters
(Mills et al., 2004; Munich Re, 2005), expected benefits read:

(2)

where d0 is the amount of damages a flooding of New Orleans would cause today. If the cost, C, of
the flood protection system is lower than the expected benefits, B, then the system should be
implemented. In spite of the difficulties already mentioned, a rough assessment of B can be made
based on current information. From historical experience (i.e. by observing hurricane frequencies
over the last century), one can evaluate the annual probability that a category 5 hurricane hits New
Orleans at about p = 1/500 (H. Saffir, quoted in Schwartz, 2005).

The Office of Management and Budget (OMB), which carries out CBA of Federal regulations in
the U.S., uses two different discount rates to analyze policy decisions (OMB, 2003; see Appendix
D, OMB Circular A-4). These two discount rates are used to assess the robustness of findings to the
choice of discount rate and to capture two approaches to CBA. First, the discount rate can be
calculated as the opportunity cost of capital, especially when strong capital reallocation is involved,
yielding a value of 7 percent in the U.S. Second, especially when the project affects consumption
patterns (e.g., fiscal changes), the discount rate can be calculated as the “social rate of pure
preference” used by the average American saver in his saving decisions, yielding a value of 3
percent. Because the New Orleans flood protection system deals with the optimal allocation of
capital, the use of the first value of 7 percent can appear a priori more appropriate. Note also that,
according to the Ramsey growth discount rule (Ramsey, 1928), the discount rate can be evaluated
by the relationship δ = ρ + α g, where ρ describes value judgments about time preferences and is
referred to as the rate of pure preference for the present; α is the elasticity of the marginal utility
and describes value judgments about the distribution of wealth; and g is the growth rate in the
considered economy. The 7 percent discount rate used by the U.S. agency is consistent with a rate
of pure preference for the present of ρ = 4 percent and an elasticity of the marginal utility of α = 1,
with an expected economic growth of the U.S. economy of g = 3 percent. The 3 percent discount
rate derived from the social rate of pure preference is consistent with a less optimistic prediction of
economic growth, at 2 percent, and a pure preference for the present of ρ = 1. It is noteworthy that
if damages are growing at the same rate as the economy, then a null pure preference for the present
(ρ = 0) would imply that the flood protection would yield infinite benefit, provided that this flood
protection system has a quasi-infinite lifetime.

2 It is assumed that protection systems have an infinite lifetime, after having checked that results were only weakly
sensitive to the protection system lifetime, chosen in a reasonable range, for the selected values of the parameters.
Indeed, as we will see, if δ ≈ g, where g is the economic growth rate (see below), the system lifetime becomes an
important variable.

Insurance and reinsurance companies (e.g., Munich Re, Swiss Re, etc…) and disaster modeling
companies (e.g., RMS, EQECAT, etc…) estimate the direct damages due to any hurricane or flood,
and their results are widely used as proxies for the overall economic cost of disasters. These
companies estimate the cost of the New Orleans flooding around $20 billion (RMS, 2005)3. Taking
into account casualties (about 1,000 people died in the flooding) raises the difficult issue of
attributing a cost to a loss of life. Because the expression “value of the human life” problematically
suggests a market in which one could buy or sell human lives, it is preferable to use the expression
“amount the public is willing to devote to reducing risk in order to save an additional life”, which
is a much more acceptable wording. Even though the value depends on the type of risk and the
probability of occurrence of the considered event, most estimates lie between $1 million and $10
million in the U.S. We will use here the commonly-used estimation of the U.S. EPA (1997) of $5
million. Given this figure, the public would be willing to pay $5 billion to avoid the human toll of
the New Orleans flood. An additional $5 billion will be added to take into account the numerous
injuries and trauma. According to these rough estimates, $30 billion seems to be a good
approximation of the New Orleans flood cost4.

Assuming that a category 4 protection system does not reduce the damages yielded by category 5
hurricanes, which is likely since there is little difference between no levees and broken levees, the
expected present benefit of a category 5 flood protection system in New Orleans can be calculated
with Eq.(2) at $1.5 billion with a 7-percent discount rate and $6 billion with a 3-percent discount
rate. Both are one order of magnitude lower than the building cost of such a system. This rough
estimate clearly rules out an upgrade of the protection system to make it able to cope with category
5 storms. It might be difficult to believe that the risk of a repetition of the devastation caused by the
category 4 hurricane Katrina is not enough to justify the implementation of the best possible
protection system. However, our CBA suggests that it is more rational from an economic point-of-
view to live the Katrina nightmare again in a more or less remote future5.

This estimate is, however, not very solidly grounded, as it does not take into account important
processes, whose impacts could be significant. In line with OMB requirements when uncertainty is
large and economic implications are in excess of $1 billion (OMB, 2003), we will now review the
parameters6 of the CBA, and propose alternative estimates. We will not, however, attribute
subjective probabilities to the various hypotheses that will be proposed and conduct a full
probabilistic analysis, as can be found, for instance, in Mastrandrea and Schneider (2005) for the
climate change issue. Indeed, current knowledge about natural disaster consequences seems still
insufficient to assess these probabilities with any confidence, and the following sections will show
how much research is still necessary.

3 Note that the losses due to the New Orleans flooding were only a fraction of the total cost of the Katrina landfall.

4 In case of a repetition of the Katrina’s scenario, a better evacuation would probably be able to avoid a large part of the
human losses and reduce this amount of damages. It has to be mentioned, however, (i) that Katrina’s track forecasts
have been very good and allowed for anticipated decisions before landfall, which is not always possible, and (ii) that an
evacuation is always subject to organisational problems and unexpected practical difficulties, making the human part of
the damages highly variable and uncertain.

5 An annual probability of 1/500 means that there is a 20 percent chance of having a category 5 hurricane hitting New
Orleans in the next 100 years, and a 33 percent chance in the next 200 years.

6 Among the necessary assumptions in the CBA, it is often useful to distinguish between the political choices that must
arise from a political process (e.g., discounting scheme), and the scientific uncertainties that can be – at least
theoretically – solved through additional research (e.g., future probability of occurrence).

Probability of occurrence

In the first CBA, historical evidence was used to assess the probability of occurrence of a category
5 hurricane landfall on New Orleans. This assessment cannot, however, be considered as robust in
this CBA. Indeed, a flood protection system has a very long lifetime. Such a long lifetime arises, of
course, from the long lifetime of infrastructures (dams, bridges, gates). But, above all, it comes
from the fact that the flood protection system will shape the city development over an even longer
time horizon. The decisions that will be taken in the next years will, therefore, constrain the flood
protection of New Orleans for at least the next century.

During this century, climate is likely to change significantly in response to human emissions of
greenhouse gases (GHG), with consequences for hurricane damages. We already detect a trend in
hurricane destructiveness: Webster et al. (2005) observed that hurricanes in the strongest categories
(4 and 5) have almost doubled in number and in proportion in 30 years; over the last 75 years,
Emanuel (2005a) detected in the North Atlantic and western North Pacific basins a strong increase
in the power-dissipation index (PDI), which is a proxy of the destructiveness of hurricanes. The
debate on the significance and persistence of these trends, however, has yet to be solved (e.g.,
Landsea, 2005; Emanuel, 2005b)

Regardless, experts expect hurricane maximum intensities to increase as temperatures rise during
the 21st century. Global climate-model experiments have suggested an increase in the relative risk
of occurrence of category-5 hurricanes under high-GHG conditions (Knutson and Tuleya, 2004). A
tropical-cyclone model run by Emanuel (2006) found that a 10 percent increase in potential
intensity, corresponding approximately to a 2°C-warming, could lead to a 65 percent increase in
PDI. Even though uncertainty remains in the amplitude of the change (Trenberth, 2005; Pielke et
al., 2005), it is noteworthy that any change in the mean characteristics of hurricanes might easily
translate into large changes in the probability of the most powerful hurricanes. As an illustration, if
the distribution shape of the hurricane PDI were unchanged, the 65-percent increase in the mean
PDI found by Emanuel (2006) would multiply by 3.8 the frequency of the 1-percent most powerful
hurricanes (see Appendix A).

Also, sea level rise and other human-induced disruptions to the Mississippi River delta (e.g.,
sediment deposition reduction) will worsen the floods associated with any hurricane falling on this
low-lying area (Burkett et al., 2003). Thus, the probability of floods currently caused only by
category 5 hurricanes might increase, as less powerful hurricanes could also produce such
devastating floods.

To take into account these two phenomena in the CBA, they will be summarized through the
assumption that the annual probability of floods currently caused by category 5 hurricanes will be
multiplied by 3.8, to reach p=1/130 years. This higher probability alone would make expected
benefits rise from $1.5 to $5.7 billion or from $6 to $22.8 billion, depending on the discount rate.

These results suggest that climate change may have an important impact on long-term hurricane
risk, even though changes in population and capital at risk will obviously be the main driver of
vulnerability during the next decades. Additionally, the large uncertainty of the future probability
of occurrence highlights one mechanism that have been disregarded in the climate change impact
literature so far, but through which climate change might be responsible for significant economic
damages in the future: climate change increases the uncertainty on parameters that impact the

design of long-term infrastructure, making them more likely to be ill-suited in the future climate. In
the present case, the risk is either to face a series of avoidable disasters in New Orleans, if the
probability of occurrence turns out to be much larger than predicted when the protection system is
designed, or to bear the sunk costs of an expensive protection system based on an overestimated
probability of occurrence.

Avoidable damages

Another major difficulty remains in the assessment of the actual damages that could be avoided
through an upgrade of the protection system. Assuming that New Orleans will be reconstructed and
that all displaced households will return to their original city (we will address this issue later in the
paper), the damages from the Katrina landfall can be used as a proxy for the damages a future flood
may cause. As mentioned earlier, however, several authors suggest that the direct costs, evaluated
by insurance companies, may be poor proxies of overall costs, especially concerning large-scale
events (Tierney, 1995; Lindell and Prater, 2003; Hallegatte and Hourcade, 2006). Indeed, direct
cost can be amplified (i) by spatial or sectoral propagation into the rest of the economic system
over the short-term (e.g., through disruptions of lifeline services7) and over the longer term (e.g.,
sectoral inflation due to demand surge, energy costs, insurance company bankruptcy, larger public
deficit, or housing prices that have second-order consequences on consumption); (ii) by responses
to the shock (e.g., loss of confidence, change in expectations, indirect consequences of inequality
deepening); (iii) by financial constraints impairing reconstruction (e.g., low-income families cannot
finance rapidly the reconstruction of their home); and (iv) by technical constraints slowing down
reconstruction (e.g., availability of skilled workers, difficulties in equipment and material
transportation, difficulties in accommodating workers). To measure the impact of these effects,
Hallegatte and Hourcade (2006) introduced in their simple economic model a parameter measuring
the ability of the economy to fund and carry out reconstruction, and derived the Disaster Economic
Amplification Ratio (DEAR), which measures the ratio between the overall economic cost and the
direct loss due to a disaster. While this ratio is less than one for small-scale disasters, DEAR is
found, using a simple model, to increase dramatically for large-scale disasters like the New Orleans
floods. This increase arises mainly from the addition to the capital replacement cost of the
production losses during the reconstruction phase. For example, if a $1 million plant was destroyed
and immediately rebuilt, the loss would be $1 million; if its reconstruction is delayed by one year,
the total loss is the sum of the replacement cost and of the value of one year of production. For
housing, the destruction of a house with a one-year delay in reconstruction has a total cost equal to
the replacement cost of the house plus the value attributed to inhabiting the house during one year.
The value of such production losses, in a broad sense, can be very high in some sectors, especially
when basic needs are at stake (housing, health, employment, etc.).

A couple of months after the Katrina landfall, initial analyses of the disaster aftermath
unfortunately confirm this intuition; because the “systemic functioning” of the whole region is now
impaired (both economically and technically), the reconstruction is dramatically slowed down and
the total cost of the event will be much larger than its direct cost. In recent publications, the RMS
research team (RMS, 2005) and the RAND Corporation (McCarthy at al., 2006) highlight the
existence of numerous “loss amplification” effects: demand surge (e.g., a 13-percent increase in
cement prices between 2005 and 2008 is forecasted); longer delay in reconstruction due to technical

7 For instance, Tierney (1995) finds that data on the consequences of the 1993 Midwest floods and the 1994 Northridge
earthquake suggest that “business properties may escape direct damage and yet suffer extensive disruption as a result
of lifeline service outages”. These short-term costs, however, are most of the time included in the assessments by the
insurance industry through “business interruption” costs.

constraints; financial constraints in reconstruction due to insufficient flood-insurance policy
coverage. These conclusions probably understate these effects, as they focus primarily on direct
costs without considering most of the second-order costs, which are not insured but are supported
by the overall community (e.g., the impact of Katrina on tourism in New Orleans and its vicinity
during future hurricane seasons, which is still unknown but might turn out to be significant; or the
cost of taking in thousands of displaced people in other cities).

Social costs of large-scale disaster also involve other dimensions than direct economic losses and
casualties, including psychological factors or political and social destabilization (see a review in
Lindell and Prater, 2003). For instance, confidence in the ability of public services to carry out
disaster relief in an efficient manner may have been undermined by this event, investors may be
reluctant to invest large amounts of money into new businesses in the affected area, and tourists
may change their favorite vacation destination if they feel endangered in New Orleans. Of course, a
comparable event occurring in the next decades would strongly amplify all these reactions, possibly
leading to a withdrawal from New Orleans, the cost of which is difficult to assess8. Also, natural
disasters have been found in the past to cause strong political and social unrest when they amplified
pre-existing tensions (e.g., hurricane Bohla in 1970, which killed almost 300,000 persons in
Pakistan without provoking an effective mobilization of the central government to provide relief,
worsened social tensions that led to the creation of Bangladesh in 1971). Even though the social
and political destabilization arising from a hurricane in the U.S. cannot but remain limited (Lindell
and Prater, 2003), the consequences of a near repetition of a Katrina-like event on the social and
economic climate of Louisiana are difficult to predict9. The cost of two Katrinas would be much
larger than twice the cost of one Katrina.

Moreover, even if long-term consequences remain limited when considering a single event, the
consequences of a distribution of events on regional economic growth can be significant
(Hallegatte and Hourcade, 2006). Indeed, as already stated by Benson and Clay (2004), poor
regions or countries affected by repeated natural disasters and suffering from a weak ability to
conduct reconstruction may prove unable to maintain the infrastructures necessary for economic
development, leading to the creation of “poverty traps.” As examples of this phenomenon, we cite
Guatemala, whose impressive series of weather catastrophes between 1997 and 2001 prevented any
development. Similarly, in Honduras, the single hurricane Michele in 2001 “put the country’s
economic development back 20 years” (Honduran Prime Minister, quoted in IFRCRCS, 2002).
Again, it is possible that disasters have, at least in some regions, long-run consequences that are
neglected in first-order estimates, through changes in potential economic growth.

For all these reasons, and given the large vulnerability of New Orleans and its vicinity (important
economic activity in sensitive sectors like tourism, transportation or energy production; low
reconstruction capacity due to a large proportion of low-income population, etc.), the extent of the
damages (80% of the city under water), and the difficulties currently met in the reconstruction
process, a conservative estimate of the actual overall cost of the New Orleans floods is at least 50%

8 In particular, a high value might be attributed to the historical and cultural content of New Orleans. In other terms,
people can decide to pay an additional price to keep living in New Orleans instead of being relocated. The choice of
this value, however, must involve a political process which is beyond the scope of this article.

9 It may also be necessary to take into account the position of the ruling government. Indeed, the government is likely
to be held directly responsible for a repetition of such an event in the same location, that would probably be considered
more unacceptable than a comparable disaster in another place (e.g., Sacramento, CA, that runs the same kind of risks
than New Orleans). This private interest may motivate an over-protection against events that have already occurred,
justifying the use of economic analysis to assess risk management projects and avoid under-optimal risk management
strategies.

larger than the insurers’ approximation based on direct losses only, to reach $45 billion. It is
difficult to assess the cost of a repetition of the Katrina scenario. Of course, the closer from the first
hit the larger the additional cost. As a best guess hypothesis, we will assume the cost of a second hit
to be, on average over the life-time of the protection system, 25% larger than the first hit, i.e.
amounting to $56 billion. Using the new values of event probability (p=1/130) and potential
damages (d0=$56 billion), the expected benefit of an upgraded protection system would be $10.7
billion with a 7 percent discount rate and $42.4 billion with a 3 percent discount rate.

The resettlement issue

We assumed in the previous section that New Orleans will be rebuilt with the same structure it had
before Katrina and that all previous inhabitants of New Orleans will return to the city, even if no
improvement of the flood protection system is undertaken. This assumption is at odds with what is
observed and what is predicted in the coming years. For instance, McCarthy et al. (2006) estimate
that the New Orleans population will be down to 272,000 in 2008, from 485,000 in 2000. If this
new population is maintained, the costs of a new flood would of course be largely reduced
compared with the 2005 one, making our assessment of avoidable damages overestimated. But it
must be remembered that the flood protection system, the design of which is currently being
discussed, will protect New Orleans for at least one century. The pertinent variable is thus the
population over the long term, not over the next decade. And the low repopulation rate predicted by
McCarthy et al. (2006) is mainly explained by the slow reconstruction pace due to short term
constraints. It provides, therefore, no estimate of the long-term repopulation of the city.

To assess the long-term repopulation, we assume that, before Katrina, the risk of hurricane was
perfectly known and that the New Orleans inhabitants had a rational behavior. We neglect here the
potentially important role of social networks (see McCarthy et al., 2006). Within this framework,
the large population of New Orleans before the storm can only be explained by comparative
advantages of the city’s location in some sectors (e.g., tourism, shipping) and by the households’
willingness-to-pay (WTP) to live there, because of environmental amenities. Both should have
more than compensated the well-known hurricane and flood risk, even in absence of improved
flood protection system. If these comparative advantages and this WTP have not been changed by
Katrina, and if basic services, infrastructures and social networks can be restored, these
assumptions mean that the New Orleans population will eventually return to its pre-Katrina level,
even in absence of improved flood protection. They also suggest that the currently observed
population reduction is more related to financial and technical constraints than to voluntary choices.

These assumptions, even though questionable, can explain the pre-storm New Orleans and allow us
to separate the design of the flood protection system from the reconstruction issue and to justify the
use of the 2005 flood data to estimate the cost of a future flood.

Countervailing risks, side-effects and reduced

Unfortunately, it is also necessary to take into account the possible side-effects implied by the
implementation of a large-scale protection system. These side-effects can yield ancillary benefits
like infrastructure improvement, as mentioned by Allenby and Fink (2005), or create or increase
other risks, referred to as countervailing risks10 by Wiener (1998), who calls for a broader
accounting of them in risk management.

10 Examples of such countervailing risks in flood management are provided by Glenn et al. (1996) or Christensen
(1997).

One cannot assess a flood protection system without taking into account moral hazard and equity
issues. A flood protection system funded through nationwide taxes, like a uniform insurance
premium, can constitute an incentive for people to settle in at-risk areas, as they do not pay for the
risk their location choice creates. Indeed, even if they prefer to live in New Orleans rather than
anywhere else, it is likely that less people will resettle in New Orleans if they think the Katrina
catastrophe can happen again than if a flood protection system makes the probability of such an
event negligible. This mechanism is potentially significant, since the large increases in population
and investments in hurricane-prone regions are responsible for most of the explosive trend in
hurricane damages observed over the last decades (Mills et al., 2004; Munich Re, 2005; Pielke et
al., 2005). It should be noticed, however, that the urbanization of vulnerable areas around New
Orleans in the past few decades does not seem to have been driven mainly by an over-protection
against hurricane floods, but rather by the trade-off carried out by low-income households, who
have high rates of preference for the present and poor access to information, between long-term
flooding risks and immediate lower housing prices.

These side-effects, however, create a paradox. We would expect an increase in the system benefits
from the fact that the protection system would allow a larger number of households to resettle in
New Orleans, where they prefer to live. It is not the case. Instead, it reduces the benefits, by
lowering the number of persons at-risk if the protection system is not built. This paradox arises
from the fact that, again, we do not take into account the comparative advantages of New Orleans
and the welfare gain or loss (or WTP) of households who would like to live in this city if they were
protected from floods. This paradox suggests that a CBA analysis of the flood protection system
taking into account countervailing risks cannot be carried out in a rigorous manner independently
of a modeling of individual location choices. Such a modeling, however, is made very difficult by
the uncertainty of household WTP, and we will have to rely on other approximations to take into
account countervailing risks in our analysis.

The importance of these side-effects will be heavily dependent on the design and practical
implementation of the protection system. In particular, huge negative consequences would certainly
result from the implementation of an ambitious flood protection system that would not be followed
by a careful long-term maintenance. In this worst case scenario, the existence of the protection
system would raise investment and population in the so-called protected area, which would not be
protected any more after a few decades of negligence, making vulnerability even larger than if no
protection was implemented in the first place. As a consequence, the implementation of a
protection system must be considered as a long-term commitment.

Also, avoiding negative outcomes from the future flood protection system requires careful design
and implementation, in order to protect already urbanized areas without steering additional
urbanization toward non-protected flood-prone locations. In this respect, the future flood protection
system in New Orleans is certainly not only a system of dams, bridges, and gates. It should also
include an important set of new regulations for future urban developments. A wisely designed flood
protection system should protect selected areas, with dams and levees, and ensure, through land-use
regulations, that investments are not attracted to non-protected areas. Hopefully, increased
experience with flood management and the high visibility of the project will foster a flood
protection plan that limits the negative effects and promotes positive ones, making the overall
consequences of these side-effects positive or, if it reveals impossible, negligible compared with
direct costs and benefits.

To be conservative, however, we will take into account the fact that a flood protection system could
increase the population and capital at risk compared with an optimal situation. To do so, we will

assume that, if no protection system is implemented, the potential damages growth rate will be
lower than nationwide economic growth, by an amount ∆g=0.5 percent, because of the influence of
hurricane risks on housing and investment location choices11. We will neglect the fact that, if the
protection system is not implemented, there is a loss of welfare for households who would move to
New Orleans if the city was protected from flood but who do not move because of the absence of
such a protection.

It means that Eq.(2) is changed into:

(3)

With this new equation, the expected benefit of an upgraded protection system would be $9.5
billion with a 7-percent discount rate and $28.4 with a 3-percent discount rate. With a project cost
of $27 billion, the decision would, therefore, be dependent on the discount rate, making it necessary
to discuss its value in more detail.

Choice of the discount rate

As already mentioned, the CBA of a flood protection system has to deal with very long time
horizons, making the value of the discount rate controversial. Indeed, there are intense debates
(Portney and Weyant, 1999) about the discount rate that should be used for environmental or long-
term issues that involve intergenerational issues. When intergenerational equity is strongly
involved, OMB suggests that discount rates between 1 and 3 percent are appropriate, since welfare
of next generations should not be discounted and only the fact that they are likely to enjoy higher
consumption levels should be taken into account. Other governments (e.g., U.K.) favor a
decreasing discount rate over time, justified by the uncertainty over future economic situations (The
Green Book, U.K. Treasury, 2003; Oxera, 2002).

Another approach is to calculate the discount rate that would make equal costs and benefits and
assess its relevance. Using our parameters (probability of occurrence p=1/130 years and potential
damages of $56 billion, rising at a growth rate 0.5 percent lower than economic growth), the
benefits of the protection system are found equal to its cost if the pure preference for the present is
lower than ρ0 = 1.1 percent. With an expected growth rate of 2 or 3 percent, it would correspond to
a discount rate of 3.1 or 4.1 percent, respectively. Since these discount rates are both greater than
the largest bound of the OMB’s discount rate range for projects involving intergenerational equity
issues (3 percent), the last set of hypotheses justifies the implementation of the $27-billion flood
protection system.

Risk aversion and damage heterogeneity

A society that would use the previous method to assess a protection system is called risk-neutral. A
risk-neutral agent is indifferent to risk, i.e. it does not see any difference between losing $1 with
certainty and having a 10% chance of losing $10, because the expected loss is the same in both
cases. Theoretically, such an agent would never pay for insurance. Regarding protection against
large-scale floods, however, there are good reasons to justify risk-averse behavior: people might

11 In other terms, the existence of the protection system is assumed to increase economic growth in the protected area
by 0.5 percent per year, making it equal to the nationwide economic growth.

indeed prefer to pay an additional amount of money (a risk premium) to avoid the risk of costly and
deadly floods.

To incorporate risk aversion, we change our assessment framework to use a utility function that
measures the welfare gain or loss that is associated with any financial gain or loss. An utility
function with risk aversion assumes that the increase in utility due to a $1 gain is smaller – in
absolute value – than the decrease in utility due to a $1 loss. As a consequence, the risk of gaining
or losing $1 with equal probability lowers the expected utility and is, therefore, equivalent to a
certain financial loss, which is referred to as the risk premium or the equivalent-certain outcome.

However, if we assume that the damages due to a hurricane landfall are perfectly shared among the
whole population of the U.S., the damage per capita is small (a few hundred U.S.$ per capita). In
such a situation, the Arrow-Lind (1970) theorem demonstrates formally why risk-aversion can be
neglected, supporting the choice of the states that consider self-insurance as a basic principle (e.g.,
France). Indeed, using an utility function with a constant relative risk aversion of one, as suggested
in Arrow (1971) for households in developed countries, the risk-premium amounts only to 0.2% of
the damages, i.e. $120 million (see appendix B).

The picture is different, however, if a substantial part of the damages impacts only by a small
fraction of the population. Indeed, when the utility function is not linear, the utility derived from
the consumption of $1 becomes lower as consumption increases. This effect represents the fact that
rich people do not gain as much from the consumption of $1 as poor people do. But, it also means
that it is not equivalent for a group of 10 people to lose $1 each or to know that one of them will
lose $10.

The consequences of these factors on the CBA analysis of a hurricane landfall can be very
significant. To assess them, we consider a hurricane landfall on New Orleans, which generates $56
billion of damages, 50 percent of which being equally supported by the whole U.S. population
(through government spending in reconstruction12, increase in insurance premium and other
propagation effects into the national economy) and 50 percent being shared only by the affected
population, which represents 0.4% of the total population, i.e. one million people. In this case,
individual losses become significant (as large as 90% of annual consumption for the affected
population) and risk aversion appears far from negligible. The loss of utility due to such an event is
found to be equivalent to the loss of utility of a hurricane that would generate $69 billion of
damages shared by the whole U.S. population, without risk-aversion (see appendix C). This value
corresponds to a risk premium of $13 billion, i.e. 23 percent of the damages. This figure can even
be considered as underestimated, because (i) low-income population being more likely to belong to
the affected population (Lindell and Prater, 2003), accounting for pre-existing income inequalities
would increase utility losses; and (ii) the actual repartition of damages is even more unequal than
we assumed, as a few people usually suffer from most of the losses (house, belongings, life
environment, but also friends and relatives). Regardless, neglecting the damage heterogeneity in the
CBA would lead to a large underestimation of the benefits from an improved protection system.

To carry out a CBA in utility terms, we must compare the utility loss due to the risk of hurricane if
no upgraded protection system is implemented with the utility loss due to the reduction of
consumption, which is needed to fund the protection system. To do so, we use the present
equivalent-certain cost of the risk of hurricane (Bc), which is the equivalent of the expected benefit,

12 Federal spending formally related to the Katrina reconstruction might even exceed the amount of damages, because
reconstruction can be used to improve regional infrastructures.

B, when utility losses are used instead of monetary costs. Bc is such that paying, immediately, the
amount Bc to build a protection system would reduce utility by the same amount than supporting
the risk of hurricane flooding. As a consequence, the protection system should be built if its cost C
is lower than the present equivalent-certain cost of the risk of hurricane (Bc). In this case, the
calculation with a pure preference for the present of 1 percent (see Appendix C) leads to a value Bc
= $35.3 billion, which makes a category 5 hurricane flood protection system economically efficient.

Conclusion

Building a flood protection system able to cope with a category 5 hurricane in New Orleans is a
huge investment, and it is wise to precisely assess its benefits before any implementation decision,
as other, less-costly projects might be more efficient to improve the population’s well-being13. One
must, however, be very careful of the underlying assumptions used in the benefit assessment.
Indeed, using probabilities derived from historical experience and direct cost estimates produced by
insurance companies lead to low assessments of benefits and rule out any additional flood
protection system. Nonetheless, making less optimistic assumptions about possible anthropogenic
increases in flood probabilities and taking into account estimates of second-order disaster costs,
public risk-aversion, and damage heterogeneity can reverse the conclusion of the CBA, even if
countervailing risks and moral hazard issues are accounted for.

These results suggest that a CBA is useful but should encompass the whole set of possible
assumptions to check its robustness. In the New Orleans case, the recommendation is highly
sensitive to these assumptions, even if the protection system is found worth building only for
pessimistic assumptions on climate change.

This analysis also shows that second-order damages and impact heterogeneity are responsible for a
large increase in welfare losses. Climate change, even though negligible during the next decades
compared to other drivers, might also be an important factor of hurricane risk over the long-term.
More generally, climate change creates an additional layer of uncertainty in infrastructure design
that increases the probability of either under-adaptation (and increased disaster risk) or over-
adaptation (and sunk costs in protection).

References

Adler, M.D., 2005. Equity Analysis and Natural Hazards Policy, AEI-Brookings Joint Center for
Regulatory Studies, Related Publication 05-30.

Allenby, B. and J. Fink, 2005. Toward Inherently Secure and Resilient Societies, Science 309, pp.
1034–1036

Arrow, K.J. and R.C. Lind, 1970. Uncertainty and the evaluation of public investment decisions,
Amer. Econ. Rev. 60, pp. 364–378

Arrow, K.J.,1971. Essays in the theory of risk-bearing, Markham Publishing, Chicago.

Arrow, K.J., M.L. Cropper, G.C. Eads, R.W. Hahn, L.B. Lave, R.G. Noll, P.R. Portney, M. Russel,
R.L. Schmalensee, V.K. Smith, R.N. Stavins, 1996. Benefit-Cost Analysis in Environmental,
Health, and Safety Regulation, American Enterprise Institute Books and Monographs

13 This is especially true if improved track forecasts, warning systems and evacuation plans can avoid human losses at
low costs.

Benson, C. and E. Clay , 2004. Understanding the economic and financial impact of natural
disasters. The International Bank for Reconstruction and Development. The World Bank,
Washington D.C.

Bohannon, J. and M. Enserink, 2005. Scientists Weigh Options for Rebuilding New Orleans,
Science 309, pp. 1808–1809

Burkett, V. R., Zilkoski, D. B., and Hart, D. A., 2003. Sea-level rise and subsidence: implications
for flooding in New Orleans, Louisiana, in Prince, K. R., and Galloway, D. L., eds., U.S.
Geological Survey Subsidence Interest Group Conference, Proceeding of the Technical Meeting,
Galveston, Texas, pp. 63–70

Christensen, J., 1997. California floods change thinking on need to tame rivers, The New-York
Times, February 4, 1997

Emanuel, K, 2005a. Increasing destructiveness of tropical cyclones over the past 30 years, Nature
436, pp.686–688

Emanuel, K., 2005b. Emanuel replies, Nature 438, E13

Emanuel, K., 2006. Climate and tropical cyclone activity : A new model downscaling approach, J.
Clim., in press

Glenn, E.P., C. Lee, R. Felger an S. Zengel, 1996. Effects of water management on the wetlands of
the Colorado River delta, Mexico, Conservation Biology 10 (4), pp. 1175–1186.

Hallegatte, S. and J.-C. Hourcade, 2005. Why economic growth matters in assessing climate
change damages: illustration on extreme events, accepted by Ecological Economics, preprint
available on http://www.centre-cired.fr/forum/article77.html?lang=en

Hahn, R.W., 2005. The economics of rebuilding cities: Reflections after Katrina, AEI-Brookings
Joint Center for Regulatory Studies, Policy matters 05-24

IFRCRCS, 2002. World disaster report 2002, focusing on reducing risk. International Federation of
Red Cross and Red Crescent Societies.

Knutson, T. R. and Tuleya, R. E., 2004. Impact of CO2-Induced Warming on Simulated Hurricane
Intensity and Precipitation: Sensitivity to the Choice of Climate Model and Convective
Parameterization, Journal of Climate 17, 3477-3495

Landsea, C.W., 2005. Meteorology: Hurricanes and global warming. Nature 438, E11-E13

Lecocq, F. and J.C. Hourcade, 2003. Equitable Provision of Long-Term Public Goods: The Role of
Negotiation Mandates, CIRED Working Paper.

Lindell, M.K. and C. S. Prater, 2003. Assessing Community Impacts of Natural Disasters, Natural
Hazards Rev. 4, 176-185

Mastrandrea, M.D. and S.H. Schneider, 2005. Probabilistic integrated assessment of “dangerous”
climate change, Science 304, pp.571–575

McCarthy, K., D.J. Peterson, N. Sastry, M. Pollard, 2006. The Repopulation of New Orleans after
Hurricane Katrina, RAND Gulf States Policy Institute

Mills, E., R.J. Roth and E. Lecomte, 2005. Availability and Affordability of Insurance Under
Climate Change: A Growing Challenge for the U.S., CERES, available on http://www.ceres.org

Munich Re, 2005. Annual Review: Natural Catastrophes 2004, Topic Geo

Office of Management and Budget, 2003. Informing regulatory decisions: report to Congress on the
costs and benefits of federal regulations and unfunded mandates on state, local, and tribal entities

OXERA, 2002. A Social Time Preference Rate for Use in Long-Term Discounting, A report for
ODPM, DFT and DEFRA.

Pielke, R.A. Jr., C. Landea, M. Mayfield, J. Laver and R. Pasch, 2005. Hurricane and Global
Warming, Bulletin of the American Meteorological Society, November 2005, pp.1571–1575

Portney, P. R. and Weyant, J.P., 1999. Discounting and Intergenerational Equity, Resources for the
Future, 202 pp.

Ramsey, F.P., 1928. A mathematical theory of saving, The Economic Journal 38 (152), p.543–559

RMS, 2005. Hurricane Katrina: Profile of a Super Cat, Lessons and Implications for Catastrophe
Risk Management, available on www.rms.com

Schneider, S.H., K. Kuntz-Duriseti, and C. Azar, 2000. Costing non-linearities, surprises and
irreversible events, Pacific and Asian Journal of Energy, 10(1):81-106.

Schwartz, J., 2005. Full flood safety in New Orleans could take billions and decades, The New-York
Times, November 29, 2005

Tierney K. J., 1995. Impacts of Recent U.S. Disasters On Businesses: The 1993 Midwest Floods
and the 1994 Northridge Earthquake, Disaster Research Center, University of Delaware

Trenberth K., 2005. Uncertainty in Hurricanes and Global Warming, Science 308, pp. 1753–1754

U.K. Treasury, 2003. Green Book, Appraisal and Evaluation in Central Government, available on
http://greenbook.treasury.gov.uk/

U.S. Environmental Protection Agency. 1997. The Benefits and Costs of the Clean Air Act: 1970–
1990. 410-R-97-002

Webster, P.J., G. J. Holland, J. A. Curry, H.-R. Chang, 2005. Changes in Tropical Cyclone
Number, Duration, and Intensity in a Warming Environment, Science 309, pp. 1844–1846

Wiener, J. B., 1998. Managing the iatrogenic risks of risk management, Risk: Health, Safety and
Environment, 9 (39), pp. 40–82

Appendix A: The link between mean PDI change and the most powerful hurricane
probability

Figure 1 shows in blue the current cumulative distribution function of the hurricane power
dissipation index (PDI), as calculated from the NHC/NOAA HURDAT database
(www.nhc.noaa.gov) between 1900 and 2004. The horizontal blue line distinguishes the 1-percent
most powerful hurricanes, i.e., in the current distribution, the hurricanes with PDI larger than
164 109 m3s-2 (vertical black line).

It has been predicted that the mean value of North Atlantic PDI could rise, between now and the
end of the century, by 50 percent, according to the predicted rise in sea surface temperature (SST)
and the observed correlation between SST and PDI (Emanuel, 2005b), or by 65 percent, according
to a hurricane intensity model (Emanuel, 2006).

If climate change increased the mean hurricane PDI by 65 percent through a shift of the distribution
to the right, the shape being unchanged, the new cumulative distribution function would be the red
line in Fig. 1. With this new distribution, 3.8 percent of the hurricanes would have a PDI larger than
164 109 m3s-2. It means that the frequency of the 1-percent most powerful hurricanes in the present
climate would be multiplied by 3.8.

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

50 100 150 200 250 300

PDI (arbitrary units)

P
ro

ba
bi

lit
y

of
e

xc
ee

di
ng

a
P

D
I t

hr
eh

ol
d

Figure 1: cumulative distribution function of hurricane PDI in the 1900-2004 climate

(in blue) and in a modified climate in which mean PDI would be
increased by 65 percent, the shape being unchanged (in red).

Appendix B: Calculation of the risk-premium

The calculation of the risk-premium is done using the following utility function, with a constant
relative risk aversion of one (see Arrow, 1971):

(B-1)

where N is the size of the population and c is the consumption per-capita (about $31,200 in 2005).
Note that we assume here that all citizens have the same consumption and that this utility function
aggregates all citizens by summing their individual utility function with the same weight. This
method is usually referred to as “one-man, one-vote”. When income inequalities are taken into
account, other aggregation rules are possible, reflecting different moral position with regard to
income distribution within a given generation (Lecocq and Hourcade, 2003) (e.g., the “one-dollar,
one-vote” aggregation rule, in which weightings are chosen so that the currently observed income
distribution is considered socially optimal).

We define the risk-neutral damages d0’ by the fact that supporting with certainty an amount of
2005-equivalent damages pd0’ every year would lower utility by the same amount than the risk of
having a hurricane that causes an amount d0 of 2005-equivalent damages with a probability p. Note
that all damages are expressed in 2005-equivalent; actual damages are supposed to rise according to
economic growth minus ∆g=0.5 percent: dn = d0 (1+g-∆g)n.

The risk-neutral damages d’0 can be calculated by the equality14:

(B-2)

This equation states that the expected utility from consumption, every year possibly reduced by
hurricane damages d0(1+g-∆g)n, with d0 = $56 billion, is equal to the utility from the expected
consumption reduced every year by the probability of occurrence p multiplied by the risk-neutral
damages d0’, also rising according to economic growth minus ∆g. Note that the rate of pure
preference for the present ρ is used to discount utility, while the discount rate δ is used to discount
monetary values.

For a disaster causing $56 billion of damages, and using the hypotheses consistent with a discount
rate of 3 percent, i.e. a growth rate of 2 percent and a pure preference for the present ρ = 1 percent,
the resolution of Eq.(B-2) gives the value of the risk-neutral damages d’0 = $56,12 billion,
corresponding to a risk-premium of $120 million.

14 Note that population is supposed to be constant.

We now define the present equivalent-certain cost of the risk of hurricane (Bc) by the equality of (i)
the utility loss due to the risk of hurricane if no upgraded protection system is implemented and (ii)
the utility loss due to the immediate payment of an amount Bc. It reads:

(B-3)

The protection system should now be implemented if its building costs are lower than the present
equivalent-certain cost of the risk of hurricane. Solving (B-3) leads to a value of Bc of $28.4 billion.

Appendix C: Calculation of the combined consequences of damage heterogeneity and risk-
aversion

The calculation of the impact of the damage heterogeneity is done using the same utility function [
Eq.(B-1) ]. If a hurricane makes landfall, and if 50% of the damages d0 are perfectly shared, while
50% of the damages are shared by one million persons, i.e. 0.4% of the population, the whole
population utility is given by:

(C-1)

where q is the proportion of affected population (here q=0.4%).

The homogenous and risk-neutral equivalent damages d0’’ are defined by the fact that supporting,
every year, with certainty and with perfect sharing among the population, an amount of damages
pd0’’, would lower the utility by the same amount than the risk of having, with a probability p, a
hurricane that would cause an amount d of damages that would be imperfectly shared among the
population.

The homogenous and risk-neutral equivalent damages d”0 is calculated using the relationship:

(C-2)

where the left-hand-side member gives the expected utility under the risk of an amount of 2005-
equivalent damages d0 that would be imperfectly shared; and the right-hand-side member gives the
utility of the consumption when an amount of 2005-equivalent damages pd0’’ is perfectly shared

every year among the whole population. For a disaster causing $56 billion of damages, the
resolution of Eq.(C-2) gives a value of the homogenous-risk-neutral damages d’’0 = $69 billion,
corresponding to a risk-premium of $13 billion. In this case, the present equivalent-certain cost of
the risk of hurricane Bc is $35.3 billion.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Mental Accounting and Consumer Choice
Thaler, Richard H
Marketing Science; Jan/Feb 2008; 27, 1; ProQuest Central
pg. 15

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Individual Decision Making and Group Decision Processes

John Payne and Arnold Wood

Arnold Wood: One of the great benefits of this group
is that you get introduced to people who are in-
credibly helpful and incredibly nice. One such
person is John Payne who has done yeoman’s ser-
vice at Duke being a dean, teacher, all those sorts
of things at one time. Yet he was nice enough to
join me in this survey of investment committees
and John is going to basically shoulder the entire
section and I’ll serve as the poster boy for this.

I have the survey here if anybody wants to look
at it.

John Payne: Thank you, Arnie.
Yesterday and today we’ve talked about issues

dealing with individuals with a focus on where
they’re not good at probability judgments, etc.
We’ve also talked about aggregates of individuals,
markets, and market behavior, either in the form of
empirical data like we’ve just seen here on histori-
cal markets situations. What I would like to do
now is focus in on another level of analysis dealing
with committee or group decisions.

In thinking about this talk, Arnie Wood and I
were faced with a challenge. There’s a lot of litera-
ture on group decision-making, but little or none on
investment committees. Arnie then gave me a num-
ber that as a psychologist is hard to conceptualize.
He said that as a rough estimation, investment com-
mittees are responsible or oversee something more
than $6 trillion in investments worldwide.

We have a situation where there’s this very im-
portant activity performed by investment commit-
tees and there seems to be almost no data on in-
vestment committee behavior. We do have a lot of
data on group behavior, just nothing or little on in-
vestment committees. Arnie and I then came up
with a plan that would have three parts to it. Part 1
is to briefly review the group literature with a spe-
cific focus on both classic and new results that we
thought were of greatest interest to investment

committee decision-making. Second, Arnie and I
wanted to collect some data on investment com-
mittees to share with you today. We’ve done a little
survey of people who are involved in investment
committees. And last, we decided to do this talk
together, combining someone who does this for a
living (Arnie) and someone who is most concerned
with understanding the psychology involved.

Now, to begin with Arnie’s view on an invest-
ment committee. Is there anyone else that serves
on an investment committee of any kind? Can I see
a show of hands here? Okay. One, two, three, four,
five, six, seven. Great. So we’ve got a lot of exper-
tise here.

If you look at the writing on committees and
group decision-making, it’s really striking. There
are two different viewpoints that you can see out in
the literature. Now, the first view is that committee
decisions are an improvement on us poor individu-
als and the idea is that none of us is as smart as all
of us. For example, we may individually have
some biases and not be able to figure out how to
estimate the probability that you’re going to have
AIDS if you get positive tests and so on, but if we
put us all together we’ll figure it out because we’ll
be smart enough as a group to do that.

And in the finance area, Merrill Lynch thought
that committees will make financial decisions and
these will be better managed. Now, I have never
been able to find the exact source of this saying,
but the old joke that a camel is a horse designed by
a committee represents the alternative view of
committee decisions. That’s unfair to the camel
because the camel has a lot of positives to it. How-
ever, consider this quote from Business Week:
“Some fund groups are successfully run by com-
mittee. However, it’s generally thought that they
diminish returns.” Also, Terrance Odean has a
wonderful paper dealing with investment clubs
that has the idea that too many cooks spoil the
profits. I should say to Terry that this particular pa-
per has caused me great grief in my family rela-
tionships. My wife is a member of an investment
club and when I shared those results with her, she
didn’t speak to me for about a week.

All right. So first question to ask Arnie Wood is,
why do funds, whether it be a museum or it be some
other organization, form an investment committee
to manage their money? What’s the idea behind it?

The Journal of Psychology and Financial Markets
2002, Vol. 3, No. 2, 94–

101

John Payne is the Joseph J. Ruvane, Jr. Professor of Manage-
ment, Professor of Psychology and Research Professor of Statistics
& Decision Sciences at The Fuqua School of Business at Duke
University.

Arnold Wood is a founding Partner, President and Chief Execu-
tive Officer of Martingale Asset Management, L.P., where he is re-
sponsible for managing the firm.

Requests for reprints should be sent to: John Payne, Fuqua School
of Business, Duke University, Box 90120, Durham, NC 27708.
Email: jpayne@mail.duke.edu

Arnold Wood: The concept is quite simple. It’s re-
ally a forum for people to share information. I
mean you can congregate a variety of skills in one
place and presumably come to better decisions,
but that’s kind of the textbook side of it. There’s
some other issues here that we’ve uncovered,
some having to do with status and so on that being
on an investment committee gives you a certain
amount of status. That, in fact, in one particular in-
stance I’m aware of, if you have important people
on your investment committee it helps your fund
raise better because these are good people and
they’ll take care of your money and they may
make lousy investment decisions for the organiza-
tion they’re supposed to be working for. But they
do raise more money than they lose and there are
the opportunity costs on the other side of it. And I
think the primary reason that we seem to think ex-
ists out there is an issue of blame. I mean these
people are agents and no one person wants to get
blamed for the responsibility of running a large
endowment or foundation. They want to share that
blame and there are a variety of reasons, some
good and some, oh, I won’t say bad, but some are
very questionable.

John Payne: When you think about it in terms of the
advantages of committees or groups, the idea seems
reasonable that none of us knows as much as all of
us. So you have different people, different back-
grounds, different experience, different knowledge
bases and if we bring them together, we’re going to
get that information shared. That is, by the way, an
interesting applied question. Who should you bring
together in a committee? One of the results we
found when we asked the question to people that
were on investment committees in terms of mem-
bership, what was the result, Arnie?

Arnold Wood: It’s an incredible number. When they
brought these people together we asked questions
like, how many minorities are on your committee.
I think in all the answers we got there were two
people who would be considered minority, either a
woman or otherwise a minority. There was no one
under 30 years of age, 90 percent of the people
were essentially white men over the age of 40, of
which most of those were 60 and beyond. So this
is a homogeneous group of people who are there
probably for other reasons, such as to talk about
things other than what they’re on the committee
for, and I’m sure you’re going to talk about a com-
mon knowledge effect.

John Payne: A piece of advice for people who are
responsible for investment committees is that if
you have a committee of, say, 16 people, large
committee, but they are related in the sense that
their judgments tend to be correlated because they
have the same backgrounds, same information and

so on. For example, if they’re correlated at about
0.3, you can do as well as that 16-person commit-
tee with a committee of four people, if those four
people are selected so that the correlation among
their judgments is much lower. Therefore, one
piece of advice, if you’re involved in putting to-
gether an investment committee, you’re much
better off putting your resources in trying to find a
small number of relatively independent judges
than you are in spending a lot of money getting a
lot of judges on a committee.

Arnold Wood: One of the less obvious reasons why
people are members of a committee is because you
put them on the investment committee, they gener-
ally are large donators to the college or wherever it
is, so they’re on there to basically—I don’t want to
say a pay-back—but to show gratitude for their
support for the organization.

John Payne: Besides information, the other good ar-
gument for committees is basically what’s called
error checking. One of us may make a flawed cog-
nitive process, but hopefully there will be some-
one else in the committee who will pick that up.
For example, you forgot to consider base rates or
you are worrying too much about sunk cost, and
we all know that sunk cost shouldn’t influence de-
cision. So the idea is you will get some error
checking.

You also may involve people into committees
simply on the value idea of incorporating different
values. So one argument would be that you might
have people on a committee who represent knowl-
edge about finance, but you also might have peo-
ple on that committee who represent knowledge
about the organization that’s being supported by
those investments and the values that they have.

An interesting question is that with all these po-
tential advantages, why do you get that negative
view of committee decisions? Why is it that too
many cooks may spoil the broth or the profits? The
argument is that there are sources of what are
called process losses. The basic idea is that when
you think about effective groups, you’ve got three
things going on. One item is what you start with in
the quality of people who are part of the commit-
tee, how good a job you’ve done at bringing in
people, how correlated they are, for example, in
their information bases, etc. That really tells you
something about the potential of the group.

Now, over and above that you would like to
have some things happen when people get to-
gether, such as face-to-face conversations and we
actually hope things are going to happen posi-
tively by having us together in this room. For in-
stance, you learn from each other. And then there
are, unfortunately, things that can perhaps work
not so well, that lead to process loss. Process

INDIVIDUAL DECISION MAKING AND GROUP DECISION PROCESSES

losses can be either cognitive losses or motiva-
tional losses. For instance, information sharing
should be one of the strengths of groups, however
there’s data that suggests it’s a place where groups
are not as good as they might be. One of Arnie’s
favorite examples of motivation loss is social loaf-
ing. That is the idea that you can hide in a commit-
tee, and maybe not work quite as hard if you’re up
there by yourself.

Finally, there are what I call social interaction
effects. These are not studied that much in the
group decision literature. One of them might be
called an illusion of effectiveness that is one thing
a lot of people seem to believe, that as a group they
are, in fact, smarter or better. And so there’s this
strong belief out there that groups are, in fact, very
effective. As part of our survey we asked people,
how correct or accurate do you think your group
is? The answer was something like 75, 80 percent
of their decisions they thought were right. We
asked them a second question that is; do you think
that groups are more or less accurate in quality de-
cisions than individuals? The majority of the re-
spondents said that the group or committee was
better. So you may get illusions of effectiveness.

We’ll also talk about some data that’s come out
recently, which is that in the discourse that hap-
pens in face-to-face meetings, there’s some evi-
dence that suggests that certain kinds of argu-
ments, perhaps representing shared values, norms
or mental models, tend to take on a special status.
And what can happen is the force of group discus-
sion reinforces that kind of shared belief.

So one thought I’ll have, and we’ll be specu-
lating about later on, is something that came up
yesterday, was the idea of new economy. The no-
tion is that to the extent that people started to be-
lieve there was a new paradigm and arguments
were being made in investment committees re-
lated to the notion of the new paradigm, the view
would be magnified by discussion. In the indi-
vidual decision literature there’s a view that a lot
of decision-making is reason based. You come
up with reasons why you prefer A over B. Well,
there’s an argument that the same thing goes on
in committees. But there are certain classes of ar-
guments that people seem to find more compel-
ling than others that build over and above what
you as an individual bring to this sort of social
interaction. I’ll give you an example of that in a
few minutes.

Voice: I was just going to say, Brad Barber and I and
Chip Heath are basically looking at that, a test of
more or less that theory with the investing clubs
and we find some evidence that the clubs seem to
buy stocks that you might have said had more ac-
ceptable attributes.

John Payne: There are certain classic arguments
that really come in here.

Voice: Yes.
Voice: I suspect, because I don’t know, I’m not on an

investment committee, that they must be orga-
nized differently, in the sense that the size of the
committee, the committee gets rewarded or not
based on the performance of the committee, wheth-
er there’s a strong chairman, the chairman rotates.
I mean these committees get organized in a lot of
different ways.

Arnold Wood: One section of the survey is on lead-
ership issues.

John Payne: Leadership and size. It’s interesting
and this is perhaps not all that surprising, the me-
dian size was 7. It was interesting and we’ll talk a
little bit about this when we talk about leadership,
the overwhelming response about what the role of
the leader was, was to seek consensus, to try to
bring people together. And a lot of strength about
majority sorts of things, which sounds really great,
but as we’ll show in some data if you’ve got a
group of people together who have biased judg-
ments, that effort to seek consensus, majority, is
just going to make that bias stronger.

Arnold Wood: 60 percent of the people who were on
committees of six or more felt they would rather
have it smaller, three to four. People liked the con-
cept of smaller committees.

John Payne: Now, before I get into the literature,
there is an important point that needs to be made
which is that you need to look at the tasks that
you’re asking groups (people) to do. There’s an
important distinction in the group literature be-
tween what are called intellectual tasks or prob-
lem-solving tasks and these are tasks where going
in we may not know what the right answer is, but if
someone comes up with the answer we all agree,
yeah, that’s right. These are sometimes called Eu-
reka problems in the literature. So you can have a
variety of people starting off with a task, you don’t
know exactly how to do it, but when someone co-
mes up with the answer most people are nodding
their heads saying yeah, that’s the right answer.

Now the interesting thing about that is that it is
really based on a shared conceptual system. So in
arithmetic we know 2 plus 2 equals 4, it’s a little
more complicated in domains like finance. But I
want to stress something here because I want to
come back to it; a shared conceptual system may
or may not be right normatively. So if you have a
shared conceptual system but it’s wrong, people
may think, yes, someone’s come up with the an-
swer, ah-ha, it’s Eureka, but it’s Eureka to the
wrong thing. Generally in this case, committees
are governed by what is called truth or truth sup-
ported wins decision schemes. So once one person

PAYNE AND WOOD

comes up an answer or one person plus another
one validates that that’s the right answer, the com-
mittee goes with it.

The other committee tasks are more judgmen-
tal or decision types of tasks, where preferences
are involved. That is, there’s no really right an-
swer, e.g., how much risk are you prepared to take
in your investment? Those tasks are generally re-
solved with some form of majority rule or strength
in proportion in support kind of process, whether
it be consensus or some variation on that. That’s
going to be important because as we talk about
how well groups do, it turns out that there’s an in-
teresting distinction between those two types of
tasks.

By the way, in looking at our investment sur-
vey, we gave people the definitions of intellectual
versus judgmental task and asked them to give us a
sense of the type of tasks they dealt with. The tasks
were investment policy, asset allocation decisions,
markets environment, manager selection review
and special issue pertaining to those specific in-
vestment decisions. Basically, they viewed these
as dealing with judgmental type of decisions
where there was no necessarily correct answer, but
there were some intellectual sorts of things and
most of the committees reported that they used
some version of majority rules.

Now, some research data. One of the arguments
is that groups will tend to do better because they’ll
check errors, etc. To illustrate, one of the classic
investment errors that people make is the sunk cost
or escalation of commitment error. To give you a
flavor of what is a typical set of results; this study
by Whyte (1993) involved giving a group of peo-
ple an investment to evaluate. The question was,
should you go ahead and invest in it or not? The
way the study had been set up was that if you just
described the forward-looking prospects of this in-
vestment, only about one-third—less than one-
third of the people thought it was worth doing, not
what you’d call an overwhelming vote for the in-
vestment option.

When they put the individuals in groups, there
was a slight tendency for the groups to go even less
with the project. However, if you then included a
of sunk cost story so you had an investment that
hadn’t worked out, but now you’ve got a chance
improve an outcome, you got a very strong indi-
vidual effect of sunk costs of the willingness to in-
vest. This is the classic escalation of commitment
effect.

What’s really interesting is look at the even
higher percentage of groups that showed the sunk
cost effect.

So here’s an example of an individual bias, es-
calation of commitment. When you have groups

make the decision, the bias doesn’t decrease, it
gets stronger. So people were much more likely to
want to go ahead with this investment with the
sunk costs story behind it than they were without
it. And the issue then became that when they did
this as a group, that effect got magnified.

Arnold Wood: This response is very intuitive to
venture capitalists. Venture capitalists come back
and said if we only had another million bucks and
we can do this, and the committee says give them
the million bucks. Let’s go.

John Payne: Groups show it even more. Now I’m
going to come back to this in a little bit, but there is
an interesting asymmetry with the group results
that I want to come back to, which is that if you
had a majority of people who were in the direction
of going with responding to the sunk costs escalat-
ing commitment, then the group just magnified
that effect. If you had a group of people who were
doing what economists teach, which is ignoring
the sunk costs, the majority tended not to do it.

But there was a little bit of an asymmetry here,
which is the likelihood that a minority would have
an impact and would actually change the opinion
of a majority was much stronger in the direction of
people who wanted to escalate commitment than
those who wanted it. And it’s to get at this point
that I’m going to stress that there may be qualities
of arguments that have an asymmetry, so an ar-
gument for why you escalate commitment was
viewed in some sense a stronger argument than ac-
tually the opposite of ignoring sunk costs.

Now, we talked an awful lot about overconfi-
dence in this meeting and I was reading a Wall
Street Journal article talking about behavioral fi-
nance and suggesting that overconfidence is per-
haps the most pervasive bias for behavioral fi-
nance and there have been some studies done
looking at groups in terms of overconfidence. One
was a study by Plous (1995) where people were
asked to estimate uncertain quantities and they
were giving confidence intervals, high-low num-
bers, and there were supposed to be 90 percent
confidence intervals on 10 items. If you look at the
number that 107 individuals got correct in this par-
ticular study, it was 3.1 out of 10; a classic over-
confidence phenomenon.

Interestingly, groups actually did better. So
they get 4.2 out of 10. Still what you might argue is
substantial overconfidence, but they’re getting
more of them right. Now, in talking about group
behavior, one common comparison is groups ver-
sus how individuals do on average. There is an-
other statistic that people who do this research
such as Gigone and Hastie (1997) suggest we
should be looking at which is what are called sta-
tistical groups. So the idea here is you take your

INDIVIDUAL DECISION MAKING AND GROUP DECISION PROCESSES

individual judgments, you form a response just by
taking, for example, the average of a number that’s
equal to the size of the group. So the group is size
four, you take four opinions and just take the aver-
age, and that gives you in some sense a baseline of
what it means just to aggregate individual opin-
ions without any face-to-face discussion at all.

An example of this that was mentioned yester-
day is consensus forecasting. A lot of consensus
forecasts, as I understand, are basically earnings
forecasts taken together and then averaged. And it
was interesting in this study that if you had just
taken these individual opinions and statistically
aggregated them, you would have gotten 7.4 out of
10 correct. That’s not an uncommon finding, that
face-to-face meetings may improve things, but
they’re still often worse than you would have been
doing without the meeting at all. Just taking the in-
dividual opinion.

Now there’s one other thing I want to stress
here. They asked these people, well, how well do
you think you’re going to do? How many do you
think you’re going to get right? What the individu-
als said was that they were going to get as individ-
uals about 5.6 out of 10 correct. Now they got 3.1,
so you can take the difference and not doing as
well as they thought they were going to be doing.
Then they asked them how do you think the group
is going to be doing. The answer, on average, was
7.5 out of 10 correct. Again, I ask you to compare
that number to the actual number of 4.2 correct
that was achieved. It’s really this idea that groups
may do better, but people often think they’re going
to do a lot better than they, in fact, do.

Now I’m going to throw out a speculation here.
If groups and our investment survey suggest this
effect, namely that it increases your confidence
and may or may not increase your accuracy equal-
ly well, then an interesting question is to take the
work that Terry’s done and others about the rela-
tionship between confidence and trading volume
and examine if there might be some relationship in
terms of the amount of trading you see by invest-
ment committees.

Voice: Just a quick question. So the response is ei-
ther 5.6 or 7.5. I mean theoretically wouldn’t we
expect the answer to be 90 percent?

John Payne: People when asked a more statistical
question tend to pool across 10 as opposed to 90
percent for individual items and you get this kind
of pattern. What’s interesting was the very strong
sense that by being part of a group, they’re going
to do that much better. I think one of the reasons
why we see so many investment committees being
formed is that we have a belief that groups are go-
ing to be just that much more better. In our invest-
ment survey, when we asked people what do you

think the probability of making a correct decision
was, and it ranged from 55 to 90 percent. Ba-
sically, 75 percent of the respondents thought that
their committee would have more confidence in
their decisions than an individual, which is consis-
tent with the prior research. As I mentioned, I
think one interesting question would be to actually
look at investment committee confidence and re-
late that to things like trading volume and so on.

Arnold Wood: And manager turnover.
John Payne: Next, when we ask people to report

how they behaved in the investment committees
and how much information sharing there was and
how much sort of persuasion, most of them re-
sponded said, we do information sharing. I happen
to believe that part of that is a socially correct re-
sponse. Sharing sounds better than persuading,
but they mentioned a lot of information sharing.
So an interesting question is how well are we shar-
ing? What are we sharing?

I’ve got two studies I want to mention in
terms of that. One is one that just recently came
out by Schulz-Hardt et al. (2000). An old and
very strong effect in the individual decision liter-
ature deals with a confirmation bias. So you have
an initial opinion, you can acquire new informa-
tion. What kind of information do you acquire,
information that supports that initial opinion or
information that might conceivably be arguing
the other side?

Well, looking at an individual, you see a lot of
evidence of what is called a confirmation bias.
That is, when people have an option to choose ei-
ther one and there’s equal numbers of confirming
or conflicting, they tend to choose more support-
ing information than conflicting information.
Again, let’s look at what happens with groups.
This bias doesn’t go away and it doesn’t get
weaker. In fact, it gets stronger. So you get as
much information being acquired, but a propor-
tionately larger percentage of that information
tends to be of a confirmation effect. And this effect
gets even stronger the larger the initial proportion
of people in favor of a position. There’s some that
suggest that if you have a strong leader of a group
who expresses an initial opinion, then you get a lot
of sort of confirming kinds of activities.

Arnold Wood: I was once an analyst who covered
the auto companies. I went through a presentation
and a chairman looked at me and he said, “Mr.
Wood this is a great report.” He then said, “My
wife drives a Ford and it’s in the shop all the time. I
don’t think we should put this on the list.” And
guess what the next person said? Gee, you know,
Ford’s having trouble with the European division
and it just got on a roll. I went out of there with my
tail between my legs and I learned about commit-

PAYNE AND WOOD

tees and leadership at that point and the role of
chairmen at that point.

David Dreman: And if there’s a tilt like we have
more recently, a very strong tilt, a lot of the com-
mittees will go the same way. They see that Gold-
man is doing this and, therefore, Shearson’s peo-
ple will probably be doing the same and First
Boston and so forth. So I think there’s an enor-
mous amount of re-enforcement of decision-mak-
ing because of the interaction among people.

Arnold Wood: Sponsorship.
John Payne: One of my favorite findings in the

group literature is called the common versus
unique knowledge issue in sharing by Stasser and
his colleagues. Basically, to just give you a sense
of this, imagine that you are trying to select a fund
manager as part of the investment committee and
you have information on different candidates for
this job. Some of that information is positive and
some of that information is negative. It is positive
in the sense of a suggestion this would be the right
candidate and negative suggesting that it wouldn’t
be. Some is just neutral and doesn’t have a strong
impact one way or the other.

In one study, they had three candidates here and
they set it up so candidate A had eight positive,
four neutral and four negative [see Table 1]. And
these were the numbers for the other two candi-
dates. So when you add up all the information, it’s
pretty clear which candidate you ought to select as
A. In fact, groups did decide on A. Then the re-
searchers set it up so they took the groups and gave
one member patterns that looked like this, whether
it be two positive to candidate David, four neutral
and four negative. For candidate Terry, a different
pattern of information would be given to a subject
but it would be a two positive for A, but it would be
a different two than were used to describe candi-
dates David and Terry. The point being that all the

information was in the heads of four people, it’s
just that some people had some parts of it and oth-
ers had other part of the information. Now if you
took all that information and you put it out on the
table, you’d be back here selecting candidate A.
What happened though was in this unshared situa-
tion, because they’d set it up so that in this case B
had four positives for most people and one nega-
tive, now that looks better, four to one for candi-
date B. So, in fact, the majority of the committee
responses were to go with candidate B.

Now, this is really interesting because the po-
tential of having all the information was there. Ev-
erybody had it and if they had fully shared it, both
the common information and the unique informa-
tion, they would have chosen A, but they didn’t.
There’s an interesting argument here in terms of
investment committees where people tend to start
off in a lot of situations sharing information that
other people have.

Now part of that is a confirmatory bias, so people
were walking in thinking B was the right answer, so
they were confirming that by giving the informa-
tion about B. But there’s some recent literature to
suggest that it’s also a sort of the enhancing effect
with people who are part of a committee. You go on
a committee and one of the things you’re trying to
do is to make yourself look good. You’ve got two
choices of information to share with the committee.
One piece of information is information that others
like and have and believe or you can share informa-
tion that’s sort of out from left field that no one else
has. Which do you do?

Well, there’s evidence that suggests that what
you do is you tend to start off, at least initially,
with information that other people already have
and it’s a form of getting status in the group. So if
everybody believes or if some people already be-
lieve Ford is not a good investment, yeah, I got
some other information about Ford that would
suggest it’s not a particularly good investment. It
makes me look better.

Let me summarize here with an old idea in
psychometrics. That is the idea that if you think
about a judgment like an earnings per share or the
prospects for Ford, you can break down the judg-
ment into three components. One component is
the truth (validity) part of your judgment regard-
ing earnings per share. Another component is the
one you see in almost all statistical models, ran-
dom error, and noise. And then third component is
another type of error; let’s call it bias, systematic
error in one direction, overconfidence, etc. What’s
interesting is that if this bias component is rela-
tively small so your main source of error is just
noise or if there is bias and it’s actually only held
by a minority of members, then it turns out groups

INDIVIDUAL DECISION MAKING AND GROUP DECISION PROCESSES

Table 1. Task—Select the Most Preferred of 3 Job
Candidates; Information—Positive, Neutral and Negative;
Groups—4 Persons With Shared or Unshared Information

Candidate

A B C

Shared information
Positive 8 4 4
Neutral 4 8 8
Negative 4 4 4
Choice % .83 .11 .06

Unshared information
Positive 2* 4 1
Neutral 4 5 8
Negative 4 1 1
Choice % .18 .62 .20

Note: Across the group all 8 pieces of information were held by the
group. Adapted from Stasser & Titus, 1985.

are going to help you and they’re going to help you
for no other reason than statistical law of large
numbers. They’re going to be canceling out this
random error.

But if you’re talking about judgment tasks
where you have reason to believe that bias is sub-
stantial and particularly if that bias is likely to be
shared by a majority or more of a committee, then
groups do not necessarily improve matters (judg-
ments). And so, Aaron, when I was talking with
you, the idea that once you get a concept like the
new economy or a new paradigm shared by a major-
ity of the group, whether it’s right or wrong, then
what happens is the group effect amplifies that bias.

So in thinking about when groups are going to
do well and when they’re not, the key is to decide
how much of what you’re going to see in judgment
is driven by this random error component versus
this bias component. If a lot of it is bias and not
random error, in the sense of noise, don’t count on
groups to fix it.

Now, I will go through this last study a little
fast, however, I think it is an important group pro-
cess result. David Schkade and others (2000) have
been involved in studies dealing with judgments
about punitive damages. In a recent study they had
over 500 juries formed of size six. A huge sample.
What the researchers did is they looked at the deci-
sion whether to award punitive damages how
much they would be. And they found a very com-
mon kind of majority rule in terms of whether to
award it.

They also found that when they asked people to
rate how much punishment the defendants should
suffer, you got what’s called a polarization effect,
so group decisions tended to either increase high
numbers or lower low numbers. But what I really
want to stress is this last result. They found what
they call a severity shift. So if they compared what
the sort of statistical results would have been
knowing what people started off thinking about
what should be the right amount to what happened
at the end of group deliberation, they found that
actually the punitive amounts people generated af-
ter the group deliberation were more severe. If the
group started off with a high number, it got a lot
higher after group deliberation.

What the authors of this study suggest is that in
this kind of domain, arguments for punishment
tended to have a rhetorical advantage. People who
were making that argument were viewed as mak-
ing stronger arguments, better arguments than peo-
ple who were arguing in the other direction. In
fact, in the study they measured how easy it would
be to make a higher or lower kind of thing and
found data supporting this sort of notion. The idea
is that while a lot of group behavior can be viewed

as just sort of a mapping of individuals into a col-
lection, majority ruled and so on. There are things
that happened in face-to-face meetings that will
have to do with arguments being shared.

Further, not all arguments are created equal.
Certain arguments tend to have an advantage. In
punitive damage award cases, arguments for pun-
ishment may have the advantage. It’s interesting in
criminal cases, there’s a leniency norm in this
country. So an argument of not punishing, not con-
victing, tends to have a special status. A question
to consider for those of you who have served on in-
vestment committees and have been part of these
meetings is whether or not there are certain classes
of arguments that may have been made, say, in the
last couple of years in the bubble and so on, that
tend to have this kind of rhetorical advantage?

So the argument is that we’re in a new para-
digm, new economy, one that sounds like it might
have a big advantage over value investing in these
kinds of settings, particularly where you’ve got
more than half the group buying into it.

Voice: Does that suggest that larger groups would be
more prone to fire under performing managers, for
example?

John Payne: Yes—let me rephrase that. I think that
groups that tend to be more likely to strive for con-
sensus are more likely to do that because that con-
sensus is going to be more based on the quality of
argument.

Arnold Wood: They’re certainly more prone to. On
the other side of this, if your allocation is 60 per-
cent to stocks and you get to 80, but you had this
10 percent bumper, people don’t go back. They get
to the 80 and they’ll stay there. It’s an amazing
thing to watch. They just won’t open the book on
what the guidelines were.

John Payne: Let me summarize because I know
we’re getting close to the ending time. Groups can
do well, particularly well, when we’re talking
about intellectual tasks where once you find the
right answer, everybody can agree that that’s true.
An interesting question is whether or not you can
occasionally get errors in intellectual tasks where
what you have is a mistake in shared mental model
of a situation. People may think that they’re doing
an intellectual task. They think they’ve found an
answer consistent with their mental model of the
situation and they go with it. They are also very
confident with the answer, but they are really mak-
ing a reasoning error.

In judgment tasks, groups show all the biases
that we have talked about. Part of the reason for
less than optimal group performance seems to be
poor sharing and confirmation bias in information
acquisition. Social loafing might also contribute to
poor information use. Again, groups don’t neces-

100

PAYNE AND WOOD

sarily mitigate attitudes. They can reinforce them.
You can get group polarization effect. Groups tend
to reinforce shared beliefs. So if there’s shared be-
liefs about leniency in criminal cases, groups will
likely exhibit that even stronger than an individ-
ual. If there’s shared beliefs about punitive dam-
ages, that becomes even stronger with groups.
Groups also feel very confident in their decisions.

Now, when I was asked to do this talk, I was
given a title, which dealt with optimizing invest-
ment decisions. So lastly, I would like to offer some
optimizing suggestions for investment committees.

First, in selecting members of an investment
committee, diversity of information and attitudes
is much more important than numbers. A four-per-
son committee properly selected to have low cor-
relation is a much better thing than a much bigger
one. Second, information sharing doesn’t happen
automatically. You need to manage that as an ac-
tive process. One of the most important things a
leader can do in an investment committee is to
manage the information-sharing process. For ex-
ample, one of the things that can help overcome
the failure to share unique information is to start
off by making sure you identify the fact that peo-
ple are in the room because they’re expected to
have differential expertise and bring that to bear
on solving investment problems.

Arnold Wood: A leader has to deputize, virtually
deputize people, say, go out of the room and come
back and tell us why this is a lousy idea. That was
done in the Bay of Pigs, by the way.

John Payne: Third, do not count on groups to cor-
rect for systematic bias. If you want to correct for
systematic bias, my suggestion is to concentrate
on training individuals using some of the tech-
niques that Baruch (Fischhoff) and others have
done to avoid some of these biases. If you try to
wait for the group to correct judgment bias, that’s
not going to happen.

Fourth, one of the things that groups can do is
they can error check, not on results, but on the de-
cision process. So if you can get the groups agree-
ing on what defines a good decision-making pro-
cess, like ignoring sunk costs or incorporating
base rate information, one thing the group mem-
bers can do with each other is to hold each other
accountable for the process by which the decision
is made.

The last point, most, I want to make is that the
process loss effects with groups are very strong
psychological effects. So you can talk about
them in a committee, you can get everybody
nodding, but then if you don’t keep working at
it, the tendency is to revert back to those biased
ways of processing information by groups. The
processes of group decision-making require con-
tinuous effort and attention for there to be
improvement.

Arnold Wood: John, thank you very much. We just
started the survey at the end of April. We sent it out
to about 130 people. It’s a rather extensive survey,
we got 13 back and I understand there are 23 sit-
ting on e-mail right now, new ones, fresh data.

101
INDIVIDUAL DECISION MAKING AND GROUP DECISION PROCESSES

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Decision biases in intertemporal choice and choice under uncertainty: Testing a common account
Chapman, Gretchen B;Weber, Bethany J
Memory & Cognition (pre-2011); Apr 2006; 34, 3; ProQuest Central
pg. 589

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Turn in your highest-quality paper
Get a qualified writer to help you with

“ M4A2: Expected Value and Consumer Choices ”

Get high-quality paper

NEW! AI matching with writer

Still stressed with your coursework?

Get quality coursework help from an expert!

Order now