Agency problems and the reporting for natural disasters : damage estimates of hurricane Katrina

(1)

1

Agency problems and the reporting for natural disasters:

Damage estimates of Hurricane Katrina

Name: Samuel Latour Student number: 10283536

Thesis supervisor: A. Sikalidis Ph.D. Date: 26-06-2017

Word count: 24,287

MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam

(2)

2

Statement of Originality

This document is written by student Samuel Latour who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

For my grandmother Riek Latour – van Bommel (July 2, 1924 – June 9, 2017), who was always very concerned with my academical development

(4)

4

Abstract

In this study, I examine the setting of a natural disaster to demonstrate that firms may use the damage reporting of the disaster to apply certain earnings management methods, more specific earnings smoothing and big bath accounting. I provide empirical evidence by analyzing the relationship between the earnings surprises and damage revisions of the firms. The damage revisions are an interesting source because the revisions indicate the fraction of the damage that firms did not report initially, and thus the amount of the initial under- or overstatement of the damage. In line with the earnings smoothing literature, I hypothesize and find that firms that issue an earnings announcement with a negative earnings surprise (bad news) are more likely to publish upward revisions of the disaster damage in subsequent years, i.e. the damage is understated in the earnings report of the disaster year. I find limited evidence for the hypotheses that in case of a positive earnings announcement (good news) firms publish downward revisions of the damage in subsequent years, and that in case of a very positive earnings announcement (very good news) the downward revision is even larger. Contrary to the earnings smoothing expectations, the big bath accounting literature imposes that in case of very bad financial results (very bad news), firms overstate costs and ‘write off’ that financial year in order to allow for better results in subsequent years. I find only limited evidence for the assumption that firms apply big bath accounting through reporting for the disaster. Also, I identify several variables that explain the observed relationships.

Key words: earnings management, earnings smoothing, big bath accounting, natural disaster, Hurricane Katrina, damage revisions, earnings surprises.

(5)

5

Suppose that you are the president of a prominent oil company in your country. Your company is about to meet analysts’ forecasted earnings for the concurrent financial year. Suddenly, your company is hit by a major disaster. You are informed that ten percent of your uninsured assets have been destroyed and you have to report this to the firm’s investors. However, the investors of the firm are ignorant. Would you report the loss immediately, or would you diffuse it across upcoming years in order to beat the earnings targets and distract investors’ attention? Most likely, you would choose the sincere option. However, history points out that not all CEO’s would.

This study serves to provide a better understanding of the agency problems that arise under reporting for natural disasters. Evidence of prior literature suggests that the reported losses of natural disasters in the US increased over the past decades (Cutter & Emrich, 2005). The National Research Council (1999, p. 1) reports: “In the United States, fatalities and injuries, property damage, and economic and social disruption resulting from natural disasters seem to have become a part of the nation's social fabric.” Therefore, this item becomes increasingly important in organizations’ financial statements. In this study, I scrutinize how in the situation of natural disasters, large information asymmetries arise. By applying several implications of agency theory, I predict that managers may misuse this stakeholder ignorance to either under report or over report the costs of natural disasters, thereby misleading stakeholders of the firm to manage the firm’s earnings. Accordingly, I attempt to answer the research question of this thesis, which is: to which extent do agency problems exist under the damage reporting of natural disasters that lead to earnings management behavior? Later on, in Chapter 2, I provide theory to support the relation between these concepts and I specify forms of earnings management that may arise, which are earnings smoothing and big bath accounting. I hypothesize that firms apply either earnings smoothing or big bath accounting. I expect the direction (upward or downward adjustment of earnings) of earnings smoothing and big bath accounting to be related to the classification of the earnings announcement (i.e. very bad news, bad news, neutral news, good news or very good news). I determine the classification of earnings announcements by calculating the earnings surprise.

(8)

8

To conduct this study, I manually collect data from 400 financial statements of firms that are affected by Hurricane Katrina. I believe that this is a proper example of a natural disaster to study because: (a) this hurricane caused a lot of damage to many organizations in the US, therefore having a high availability of data; (b) the US is known to have reliable accounting standards and consistent application of these, therefore enabling comparison across firms and industries; (c) the fact that the hurricane occurred in 2005 implies that the effects and accounting in subsequent years have passed and can be observed.

As a data source, I observe the damage revisions that firms publish in years subsequent to Hurricane Katrina. In these revisions, firms either adjust the damage upward or downward. Accordingly, I find some evidence (in case of bad news) that may indicate that firms apply earnings smoothing. For the big bath accounting hypothesis, I find no evidence. In an attempt to explain the results, I find that the strength of firms’ corporate governance explains part of the relationship between the earnings announcement category and the amount of the damage revision. More specific, the results indicate that both the size of firms’ board and the percentage of directors in the board that reached the retirement age are positively associated with the damage revision amount.

As there is relatively little prior literature on the topic of reporting for natural disasters, I believe that this study makes important contributions to this research field. Regarding this topic I am only aware of a very recent paper from Michels (in press) and a paper from Bonetti et al (2015), who study reporting for natural disasters in a very different context1_{. This allows to make}

several contributions. First of all, this study makes a theoretical contribution by studying the implications of agency theory in the setting of a natural disaster. Certainly, agency theory is a research field that has extensively been studied in prior literature. However, linking this theory to the unexplored concept of reporting for natural disasters is a unique research topic. It results from this study that these concepts are definitely interesting to combine.

1 Michels (in press) studies the different effects of disclosure and recognition of financial statement information in the setting of a natural disaster. Bonetti et al (2015) study firms’ environmental disclosures subsequent to a natural disaster to prove that firms that disclose more environmental information experience a less severe increase in the cost of capital due to the disaster.

(9)

9

Second, this study delivers practical contributions to the financial reporting expertise. Currently, a firm’s stakeholders do not have a lot of guidance when they are confronted with accounting estimates of natural disasters. This study shows in practice that stakeholders should be aware of potential agency problems that may lead to earnings management behavior and this study provides stakeholders with an overall framework to judge a firm’s estimations of the impact of a natural disaster and roughly adjust these estimates, thereby giving them a stronger position when such agency problems arise.

The remainder of this paper is organized as follows. In Chapter 2, I give the background for this study by explaining the damage reporting process for natural disasters, and Hurricane Katrina specifically. Also, I provide the background for the applicable theories by describing agency theory and earnings management literature. I finish this chapter by stating the expectations and corresponding hypotheses of the study. In Chapter 3, I describe the methodology of this paper by explaining the sample selection and data collection processes and I define and explain the variables in the research model. In the following chapter, I derive the results for the different hypotheses and I come up with possible explanations. In the final section of this paper, Chapter 5, I conclude on this paper.

(10)

10

2. Literature

In this section, the background for the study is illustrated. I discuss the damage reporting of natural disasters in general and the accounting implications of disasters. Thereafter, the agency theory and the earnings management theory are introduced and the concept of natural disasters is linked to these theories. Eventually, this leads to the development of expectations and the formulation of hypotheses at the end of this section.

2.1 Background

In discussing the background for this study, first of all, I describe natural disasters in general and the difficulties that arise when accounting for them, in order to provide a context for the following part of this study. Subsequently I describe the process of establishing accounting estimates. Third, I describe the effects of Hurricane Katrina and the accounting implications of Hurricane Katrina, which I examine to conduct this study.

2.1.1 Natural disasters

When an organization is hit by a natural disaster, such as an earthquake, a storm or a flooding, this may cause significant damage to the organization’s assets and disrupt its operations. The damage that is caused by natural disasters is stated in the financial statements of organizations. Because the reporting of costs should be matched to the period in which the event occurred, the organization should recognize the costs of the natural disaster for the entire amount in the concurrent financial year (Ernst & Young, 2012). This concerns making significant accounting estimates in the year that the disaster occurred. Many factors are involved in this process.

For instance, the disaster may lead to asset impairments, disruptions of operations, insurance claims and lawsuits with other involved parties (Ernst & Young, 2012). For some of these factors, it is difficult to make reliable accounting estimates in the year that the disaster occurred. One can think of insurance claims that are settled after many years or the leaking of chemicals due to the disaster, leading to long-term ground contamination. For these factors, the organization’s total costs are hard to estimate in the period immediately after the disaster. In the following subparagraphs, I summarize different kinds of events that may occur as a result of a

(11)

11

natural disaster and I discuss the corresponding accounting implications based on a technical reporting publication from Ernst & Young (2012). In the publication, these events are marked as specific complications that should be considered in case of a natural disaster and are interesting to describe for this study to illustrate the accounting irregularities that arise.

2.1.1.1 Asset impairments

A company may show a loss due to a natural disaster either through the impairment of an asset or the incurrence of a liability. Indications of asset impairments may follow directly or indirectly from a natural disaster. For instance, impairment of a damaged factory is a direct result of the disaster. An indirect result of the disaster can be damage to infrastructure surrounding the factory, which deteriorates the geographic position of the factory and leads to an impairment because of a lower value in use.

An impairment analysis is required when indicators exist. For a natural disaster, these could be: a large decrease in the market price of the asset, a modification in the way the asset is utilized, a change in the asset’s physical condition, a new expectation that implies that the asset is likely to be sold or disposed long before it reaches its initially estimated useful life. The complete destruction of an asset results in a write-off of the asset instead of impairment. For several specific asset categories, such as inventory and debt and equity securities, more specific impairment guidance exists.

2.1.1.2 Insurance claims

Assets that are impacted by a natural disaster are often insured. When assets are affected and an insurance claim is approved, this leads to the conversion of nonmonetary assets into monetary assets. Therefore, even though assets are insured, the effects of the disaster should still be recognized by writing off on the asset and recording insurance proceeds. The general rule is that recoveries may be accounted for when they are probable of receipt. Thus, the accounting for insurance recoveries becomes more complicated when the insurance coverage is not paid out immediately but in a future period and the recoverable amount cannot be determined reliably before it is paid out. In that situation, an organization is required to incur a loss for the entire

(12)

12

amount of the damage initially, regardless of whether the affected assets are insured. The insurance proceeds should in that case be recorded upon receipt from the insurance company. Also, insurance recoveries may not be recorded before the corresponding losses have been recognized.

Another difficult situation regarding insurance recoveries arises when a company expects to receive insurance proceeds in excess of the recognized loss due to the disaster. In that case, the company should record a gain due to the recoveries. However, this gain may only be recognized as soon as all contingencies relating to the insurance claim have been resolved. Finally, it is important to note that insurance recoveries are accounted for on a property-by-property basis instead of a portfolio approach. This results in a high amount of journal entries being recorded due to different insurance recoveries, and the total expected recoverable insurance amount consists of many different sub-balances.

2.1.1.3 Business interruptions

The impact of a natural disaster may in many cases lead to interruptions of business operations. Accordingly, besides incurring costs to restore operating facilities, firms are affected by losing revenues due to the inability to continue production operations. First of all, restoration costs may be hard to predict in case of a complicated restoration process. The restoration of some plants may take years. Additionally, for these restoration operations, firms often rely on the efforts of third parties and therefore are not always able to come up with their own accounting estimates. Second, firms commonly have insurance policies in place that cover part of the incurred restoration costs and the lost revenues. However, there is a wide variety of policies and these require careful analysis, as the damage that is covered is often not clearly specified. For instance, many policies cover temporary relocation costs, and others cover lost revenue or operating margins which are measured over a long period and is compared with earlier years. If revenues recover during the period that is measured, then no insurance recoveries are paid out by the insurance company. Therefore, it is often difficult to predict what the recovered amount as covered by the insurance policy will be in case of business interruptions.

(13)

13

2.1.1.4 Exit or disposal of projects or assets

A natural disaster may lead to the sale or abandonment of assets. Exit activities could include: a significant reorganization, the sale or termination of a business unit and the closure or relocation of a regional business location. Even though an event does not have a material impact on an entity’s business, it can be considered an exit activity. Disposal activities relate to the disposal of assets due to sale or for other reasons.

Exit or disposal activities can immediately lead to high costs for firms. For example, costs relating to employee termination benefits, contract termination costs and closing or relocation costs can be involved. In the context of a natural disaster, one can think of a destroyed oil platform that is abandoned. Firms facing exit or disposal costs due to a natural disaster should recognize a liability at fair value to account for the costs they will incur later on, i.e. the costs are most often accounted for as soon as an exit or disposal decision is made.

2.1.2 Accounting estimates

Now that the different complex events that may occur as a result of a natural disaster are described, it is important to develop an understanding of the implications of accounting estimates. In this section, based on accounting and auditing standards, I attempt to answer several questions with respect to accounting estimates, such as: when is it necessary to make accounting estimates? Who is responsible for making accounting estimates? How are accounting estimates established? How do auditors account for accounting estimates in their audit procedures? Is it possible to verify the accuracy of an accounting estimate?

Accounting estimates in financial statements measure the effects of past business transactions or events, or the present status of an asset or liability. In SAS 57, which gives the US GAAP audit approach for accounting estimates, an accounting estimate is defined as “an approximation of a financial statement element, item, or account.” It can be necessary to include an accounting estimate in an organization’s financial statements either because the measurement or valuation of accounts is uncertain as it relies on future events, or because the relevant data for accounts cannot be obtained on a timely, cost-effective basis.

(14)

14

According to SAS 57, an organization’s management is responsible for making accounting estimates or establishing a process for the preparing of accounting estimates. Estimates can be based on both subjective and objective factors. Therefore, management’s judgment is commonly required to make an estimate. SAS 57 states that “management's judgment is normally based on its knowledge and experience about past and current events and its assumptions about conditions it expects to exist and courses of action it expects to take”.

By decomposing the previous sentence it draws that management’s judgment is normally based on any of the following three sources:

(1) Knowledge and experience about past and current events; (2) Assumptions about conditions management expects to exist; (3) Courses of action management expects to take.

As all three sources of management’s judgment consider either specific knowledge of the manager or the manager’s expectations, this implies that a high degree of managerial discretion is involved in the process of forming accounting estimates. Furthermore, based on this information, it seems that management’s judgments are hard to verify by individuals who do not possess management’s business knowledge. This is supported by a study from the National Research Council (1999), which is conducted to provide an accounting framework for assessing the losses of natural disasters. The National Research Council (1999, p. 27) states: “In addition to the lack of a comprehensive data base, there exists no standardized estimation technique or framework for compiling loss estimates from individual disasters. Most estimates are ad hoc, consisting of those losses that were significant in a particular event. As a result, the range of loss estimates of a natural disaster tends to vary widely, sometimes as much as 10-fold.” This statement affirms that management has a lot of discretion to make accounting estimates.

SAS 57 also states how auditors can evaluate the reasonableness of accounting estimates. However, this is not very specific: “Accordingly, when planning and performing procedures to evaluate accounting estimates, the auditor should consider, with an attitude of professional skepticism, both the subjective and objective factors.” This indicates that there is no predefined manner to respond to accounting estimates as an auditor, and what kind of evidence to obtain. Further, SAS 57 states that an organization’s internal control can be capable of reducing the

(15)

15

likelihood that material misstatements occur in accounting estimates. For instance, the presence of qualified personnel, adequate review procedures and relevant and reliable data are considered such components of internal control.

Applying the discussed implications of SAS 57 to accounting estimates of natural disasters, I illustrated that the damage estimations of natural disasters concern a lot of managerial discretion. By applying their judgment, managers are enabled to rely on their personal knowledge and expectations. Also, I illustrated that auditors and investors are not perfectly capable of verifying management’s judgments, as they often do not possess the same personal information as the organization’s management. Therefore, both management’s discretion is high and verifiability by other parties is low for the damage reporting of natural disasters. Additionally, I illustrated in the previous paragraph that the cost accounting of natural disasters is very complex. I come back to these observations in Chapter 2.2.

2.1.3 Hurricane Katrina

“A tropical depression was observed on Tuesday, August 23, becoming a tropical storm by Thursday. By Friday, this depression had become serious enough that the Governors of Mississippi and Louisiana declared states of emergency. National Weather Service forecasts changed predictions, first saying that the hurricane was heading to New Orleans at 11 a.m. on Friday. By 4 p.m. the storm was predicted to hit the Mississippi Coast. By 4 a.m. on Saturday New Orleans was again expected to be hit. On that day voluntary evacuations began in Louisiana, President Bush declared a state of emergency and FEMA and state emergency responders began 24 hour operations. By 7 p.m., the National Weather Service warned that levees could be topped in New Orleans, causing catastrophic flooding” (International Risk Governance Council, 2009, p. 2).

In this study, I focus on Hurricane Katrina to examine a specific case of a natural disaster. According to CNN (2006), Hurricane Katrina is the “costliest hurricane in US history”, with a total estimated damage of USD 108 billion. The hurricane, which occurred in August 2005, caused the deaths of over 1,800 American citizens and made over one million people lose their homes (International Risk Governance Council, 2009; CNN, 2006). Hurricane Katrina hit the southeast

(16)

16

states of the US; more specific the states Louisiana, Florida, Texas, Alabama and Mississippi. Besides causing strong windstorms, the hurricane led to severe flooding and tornadoes.

The hurricane impacted several industries in varying ways. The oil industry was impacted through large oil spills and destruction of facilities. The resulting oil shortages led to a large increase of the oil price, therefore impacting other industries as well (Herman, 2006). Also, the utilities industry suffered heavily from the hurricane, due to the destruction of 90% of the local utility networks resulting in disruptions of operations (International Risk Governance Council, 2009). This in turn impacted other industries, as power shortages affected many organizations. Another case of an industry that was particularly hit hard by Katrina is the insurance sector. According to The Economist (2005), successful insurance claims from organizations due to Hurricane Katrina amounted to approximately USD 40-60 billion, thereby leading to large losses for insurance firms.

A case study that was issued by the International Risk Governance Council (IRGC) in 2009 points out that the impact of Hurricane Katrina could have been mitigated partly, were it not that many human failures occurred. First of all, it turned out that governmental institutes made a poor assessment of the impact of the hurricane by significantly underestimating its effects. Prior to landfall, Federal institutes lacked urgency and treated Katrina as if it was a storm like any other that occurs frequently in the southeast of the US. Even after landfall, no real progress was made with respect to coordination and communication. The involved organizations were out of touch with the area that was affected by the hurricane and their knowledge was lagging behind media reports of the disaster. This was partly due to the large impacted area of approximately 90.000 square miles, which made it difficult to fully comprehend all the actors involved.

Another cause of the large impact was the poor preparation prior to Hurricane Katrina. Although the dangers of such a large hurricane had been long-anticipated, there was no operational response plan due to insufficient funding. Particularly for New Orleans, which is built mostly below sea level, this plan was essential. Finally, the IRGC report states that in years prior to Hurricane Katrina the Federal Emergency Management Agency (FEMA), which is responsible for crisis management and coordination, was significantly cut in its governmental funding. This reduced the operating effectiveness of this organization and its ability to provide support.

(17)

17

Concluding, the impact of Hurricane Katrina could have been substantially smaller were it not that there were issues with respect to the assessment of the impact and the sense of urgency, the communication and coordination, the preparation before Katrina, and the FEMA’s funding. Several of these issues were addressed in the Post Katrina Emergency Management Reform Act (PKEMRA), which was issued after Hurricane Katrina to implement several structural changes for a better response to future disasters.

2.1.4 Hurricane Katrina – Accounting classification

Reporting for natural disasters can either occur in the normal income statement or in other comprehensive income. Under other comprehensive income (OCI), the costs are reported as ‘unusual items’ or ‘extraordinary events’. Reporting costs in OCI implies that the costs are not included in net income, which is a signal to stakeholders that the costs are not incurred due to the normal operations of the firm (Fairfield et al, 1996).

To report an event as an unusual item or extraordinary event under US GAAP, two criteria have to be met: the event should be infrequent in occurrence and unusual in nature (Ernst & Young, 2012). The FASB concluded that for Hurricane Katrina, these were not met:

“ "As tragic as hurricanes and other natural disasters are for everyone affected, unfortunately every year many businesses across the country are affected by these types of events and thus they do not represent an unusual and infrequent occurrence to businesses or to insurers," wrote FASB spokesman Gerard Carney in a statement responding to a question” (Gullapalli, 2005, p. 1).

Moreover, Gullapalli (2005) suggests that classification of an event as ‘extraordinary’ is very uncommon, because the FASB often prefers to avoid this type of classification. The reasoning behind this is that reporting in the normal income statement is most useful to stakeholders, as there is no need to separate between the effects of the natural disaster and the effects of normal business activities. It is often difficult to verify whether the costs incurred under extraordinary events are accurate and whether no effects of the event (e.g. impairments, plant shutdowns) have been reported in the ordinary income statement, and vice versa. This statement is supported by

(18)

18

Elliott and Hanna (1996), who study accounting write-offs and the information content of earnings. Elliott and Hanna (1996) state that:

“Unusual items may obscure the information contained in reported earnings numbers. For example, firms might transfer normal components of operating expenses into the special item and thereby artificially increase both current and future earnings before special items. Such reporting may complicate assessment of core earnings, recurring earnings, and other valuation constructs” (p. 136).

2.2 Agency theory

In this section, I first discuss the agency problems that arise under the damage reporting of natural disasters. After discussing the agency theory I identify the actors that are involved in the principal-agent relationship.

2.2.1 Agency theory

As illustrated in paragraphs 2.1.1 and 2.1.2, accounting for the costs of natural disaster is a very complex process that involves a high degree of managerial discretion and low verifiability by other parties. The complicated damage estimates lead to large information asymmetries between the firm’s top management and its stakeholders. In most cases stakeholders can only rely on the information as disclosed by the firm in its financial statements and newsfeeds, as they do not have many alternative information sources to rely on. Under circumstances where a lot of information asymmetries are involved between principal and agent, so called agency problems arise.

Agency theory has been applied in many different fields of research, for instance in economics, finance, accounting, political science, organizational behavior and sociology studies. Agency theory seeks to explain the situation in which one party, called ‘the principal’, delegates work to another party, called ‘the agent’. In this situation, the agent performs the work, but the principal bears most of the risks related to the results (Eisenhardt, 1989). In this principal-agent relationship, there is often a separation between ownership and control, and both the principal and agent strive to maximize their utility in this relationship. As both parties intend to maximize

(19)

19

utility, the agency theory proposes that the agent will not always behave in line with the interests of the principal (Jensen & Meckling, 1976).

In the agency theory literature, two main agency problems are identified (Eisenhardt, 1989). The first agency problem arises when there is a conflict between the goals of the principal and agent or when it is difficult for the principal to monitor the actions of the agent. This agency problem is characterized by the inability of the principal to verify whether the agent has behaved appropriately. The second agency problem is the result of risk sharing that arises when the principle and the agent have a different risk appetite. A consequence of this problem is that the principle and agent often prefer different actions and decisions because of their different risk attitudes (Eisenhardt, 1989). An important example of an agency problem is given by Jensen & Meckling (1976), who examine the equity costs of agency. A CEO who owns 5% of the shares of a company would benefit from all perks that he consumes in his role as CEO of the firm, as only 5% of the costs and benefits of the firm are on his expense. However, the CEO seizes 100% of the benefits from the consumption of perks. This may lead to undesired perk consumption by the CEO.

To deal with the two problems, agency theory studies strive to identify the most efficient contract governing the principal-agent relationship. According to Eisenhardt (1989), this concerns evaluating assumptions about self-interest, risk aversion, bounded rationality, conflict of goals among organizational members and the marketability of information. In identifying the most efficient contract, there are two general directions. A contract can be either input-oriented (behavior) or outcome-oriented (result). Where an input-oriented contract focuses on salaries and hierarchical governance, an outcome-oriented contract focuses on commissions, stock options and market governance (Eisenhardt, 1989). An input-oriented contract is particularly useful in a situation wherein the principal is capable of identifying the right actions for the agent to take and monitoring whether the agent complies with the desirable actions. An outcome-oriented contract, on the contrary, is applied in situations where the principal is not capable of identifying desired actions and monitoring compliance with these actions by the agent. Instead, in an outcome-oriented contract the principal identifies several performance measures and uses incentives to induce the agent to direct performance along with the performance measures

(20)

20

(Merchant & van der Stede, 2012). In this situation, the agent has the autonomy to select the actions that lead to the desired result. Through the performance measure outcome, the principal infers whether the agent has taken the appropriate actions in an ex post fashion. However, because the performance measures can be influenced by many external factors, the agent is at risk under this contract. These external factors are referred to as ‘noise’ and consist of factors over which the agent has no control (Merchant & van der Stede, 2012). The noise that is generated may drive the agent away from the desired performance outcomes. Therefore, both input- and outcome-oriented contracts have some deficiencies, and often the two contract types are used in combination to mitigate the risks of the deficiencies.

In this section, I described several general aspects of the agency theory that are important to understand the foundation of earnings management practices2_{, which I discuss in Chapter 2.3.} In the next section, I identify the parties that are involved in the principal-agent relationship under the damage reporting of natural disasters.

2.2.2 Parties involved in principal-agent relationship

Applying the agency theory to the case of reporting for natural disasters, several implications can be derived. The principal-agent relationship consists of the investors, who together make up the principal; and the board of directors, which is the agent in this relationship. The board of directors is given several incentives to achieve the desired performance outcome. As most top managers receive both a fixed salary and several performance-tied bonuses, the agent is incentivized both with input- and outcome-oriented contracts. In the situation of a large natural disaster, the performance measures as included in the outcome-oriented contract are likely to be affected by the noise from the disaster, due to the effect on share prices and thus stock options.

In the event of a natural disaster, several external parties are involved that take part in the principal-agent relationship. These parties account for the monitoring function that gives the principal a better overview of the actions of the agent. The first party to fulfill the monitoring

2 For a more comprehensive overview of the agency theory, please refer to Eisenhardt (1989) and Jensen and Meckling (1976).

(21)

21

function is the auditor. As stated earlier, auditors are responsible for evaluating the reasonableness of accounting estimates with regard to natural disasters. Based on SAS 57, auditors have three specific objectives in doing so. Auditors should obtain sufficient appropriate audit evidence to provide reasonable assurance that: all material accounting estimates have been developed; those estimates are reasonable in the circumstances; and that accounting estimates are presented according to relevant accounting principles and the accounting estimates are properly disclosed. Further, the auditor is responsible for assessing the strength of the internal control, which according to SAS 57 is capable of reducing the likelihood that material misstatements occur in accounting estimates.

Second, to comply with their SEC filing requirements, firms disclose information about natural disaster effects in their 10-Q and 10-K reports. If the SEC receives a notice (e.g. from an analyst, surveillance organization or whistleblower) that there may be issues with the disclosed information, this can lead to investigation procedures. The information about natural disasters that is part of the financial statements can be subject to such investigations. The third party in the monitoring function is the FEMA, an agency that has been discussed earlier. The FEMA is responsible for disaster recovery aid, both for individuals and organizations. In case of a disaster, organizations can apply for a “low interest disaster loan” (http://fema.gov/). In order to receive a loan, the FEMA needs to verify whether the reported damage is accurate. Thus, the FEMA also fulfills a monitoring role in the principal-agent relationship.

Despite the presence of different monitoring parties, managers still have the opportunity to manipulate information that is disclosed to investors. I discuss this in more detail later on.

2.3 Earnings management

In this section I discuss the earnings management possibilities that arise in case of agency problems, which I narrow down to earnings management possibilities under reporting for natural disasters. Eventually, I discuss the two earnings management methods that are examined in this study. Before I discuss specific earnings management methods, I first provide a general context for the concept of earnings management.

(22)

22

According to Kothari et al (2008), a result of agency problems is that top managers engage in earnings management activities. By using earnings management practices, managers (often illegally) manipulate earnings either through reporting or decision making. By manipulating earnings through reporting, earnings are managed ex-post by displaying the economic events in a way that is favorable to top management. Through decision making, top management is able to manage earnings in an ex-ante fashion by manipulating the actual economic events. Earnings management activities are applied by top management, e.g. to meet targets or forecasts that are part of their performance measures or to meet debt covenants. Therefore, earnings management activities are a negative consequence of agency problems.

2.3.1 How is top management able to manage earnings?

First of all, it is important to mention that earnings management possibilities arise in situations where management’s judgment is required (Dechow & Skinner, 2000), a concept that has been explained earlier on in this study. The reasoning behind this is that earnings management is not used to present ‘alternative facts’, as this would be demonstrable fraud. Instead, earnings management is used in situations where management has to make an estimate, which is much harder to challenge by accountants and regulators.

Dechow and Skinner (2000) state that certain forms of earnings management are hard to distinguish from appropriate accrual accounting choices and they argue that accrual accounting is actually closely related to earnings management. Accrual accounting is used to record the financial effects of transactions and other events that have consequences for an entity in the periods in which the events occur, rather than only in the periods in which the cash is received or paid by the entity (Dechow & Skinner, 2000). Thus, Dechow and Skinner (2000, p. 237) argue, “the principal goal of accrual accounting is to help investors assess the entity's economic performance during a period through the use of basic accounting principles such as revenue recognition and matching”. As a result, reported earnings are often smoother than the underlying cash flows and earnings provide more accurate information to investors than cash flows (Dechow & Skinner, 2000). However, when does the appropriate exercise of managerial discretion in order to guide investors become a practice of earnings management?

(23)

23

2.3.2 Is earnings management considered illegal?

According to Dechow and Skinner (2000), earnings management can occur either within the spectrum of accounting choices of GAAP or it can be a violation of GAAP (although it is not said that a violation of GAAP is necessarily an indication of earnings management). Even though earnings management may occur within the boundaries of GAAP, this commonly has the same adverse consequences for firms and top managers as financial fraud. For instance, the practice of earnings management can lead to SEC enforcement activities in the same way as financial fraud. Therefore, even though GAAP is not violated, earnings management is treated in the same manner as fraud because it concerns intentional manipulation of financial figures (Dechow & Skinner, 2000). Accordingly, earnings management is per definition considered illegal. However, as Dechow and Skinner (2000, p. 239) argue:

“Perhaps the main point to be made here is that there is a clear conceptual distinction between fraudulent accounting practices (that clearly demonstrate intent to deceive) and those judgments and estimates that fall within GAAP and which may comprise earnings management depending on managerial intent. However, in the case of the latter types of choice it would, in many cases seem difficult, absent some objective evidence of intent, to distinguish earnings management from the legitimate exercise of accounting discretion.“

2.3.3 Examples of earnings management practices

Schroeder (2005) mentions several reporting and decision making earnings management techniques that are commonly used by top managers. For instance, top managers can take a bath by overstating costs once to reduce future expenses, abuse the materiality concept by recording small errors that are ignored due to the materiality concept, or engage in earnings smoothing activities by building up “cookie jar” reserves (Schroeder, 2005). For other earnings management practices, please refer to Schroeder (2005).

(24)

24

2.3.4 Earnings management and reporting for natural disasters

In the case of reporting for natural disasters, principally the event has already occurred and the top management of a firm does not have any influence on the event, as this is triggered by an externality. Thus, managers can only manage earnings by misusing management’s judgment through reporting (ex-post) and not through decision making (ex-ante). The situation is that top managers have to make an estimation of the damage that is caused by the disaster, which is often established by using management’s judgment (Kirschenheiter & Melumad, 2001). By either under reporting or over reporting the damage of the disaster in the company’s financial statements, top management is able to hide several of the effects of the natural disaster initially and disclose the true effects later on. According to different researchers (Kirschenheiter & Melumad, (2001); Kothari et al (2008)) the practice of under reporting or over reporting income can be considered beneficial to firms’ top management. Kirschenheiter & Melumad (2001), who study a financial reporting model where investors deduce the precision of reported earnings, for example argue that managers are rewarded for providing smoother earnings. Also, Khan (2005) and Eisenhardt (1989) state that managers are known to behave in their own interests and report especially self-serving information. Therefore, because managers have discretion in their financial reporting estimates (Dechow & Skinner, 2000; Schroeder, 2005), because managers are proved to benefit from the manipulation of earnings (Kirschenheiter & Melumad, 2001; Kothari et al, 2008), and because managers are assumed to act in their own interest (Khan, 2005; Eisenhardt, 1989), managers are likely to misuse the subjectivity that they are granted due to event of a natural disaster. Referring to the earnings management techniques as mentioned by Schroeder (2005), techniques that can be applied to under report or over report the damage of natural disasters are earnings smoothing and big bath accounting. Therefore, I examine these to measure the extent to which earnings management is applied under reporting for natural disasters. I discuss these two concepts in the following sections.

(25)

25

2.3.5 Earnings smoothing

Trueman and Titman (1988) define earnings smoothing as:

“The manager shifting the recognition of some of the firm's income, if there is that flexibility within the firm, from the second period to the first (the first period to the second), whenever the first period's economic earnings are less than (greater than) the expected per period economic earnings” (p. 129).

As stated before, earnings smoothing opportunities arise in situations with many accounting accruals and situations in which management has to apply judgment to provide estimates. Earnings smoothing then regards the shifting of income through periods of time in the short run to create a smoother income stream that suits investors’ and analysts’ expectations, as it is perceived that a smooth income stream is beneficial for a firm’s share price (Trueman & Titman, 1988; Kirschenheiter & Melumad, 2001). For example, this can be done through aggressive accounting by understating bad debts provisions and understatements of asset write-offs. Equally, it occurs that an organization uses conservative accounting to smooth earnings in the opposite direction.

The decision to smooth earnings depends on analysts’ expectations. Managers do not want to deviate from these expectations excessively, as this deteriorates the smoothness of the income stream (Trueman & Titman, 1988; Kirschenheiter & Melumad, 2001). Therefore, managers are aimed at establishing a small earnings surprise (also called unexpected earnings), either positive or negative. In case of a positive earnings surprise, an organization is regarded to release good news. In line with Trueman and Titman (1988), a manager that applies earnings smoothing would then decrease reported earnings to shift towards analysts’ expectations. This will give him the opportunity to also meet analysts’ forecasts in future years3_{. Vice versa, in case of a negative} earnings surprise an organization is regarded to release bad news. A manager that applies earnings smoothing would then increase reported earnings to suit expectations.

3 If the manager would excessively outperform analysts’ forecasts, then the forecasts in the next year will be based on the good performance, and most likely the manager will not be able to outperform the forecast again.

(26)

26

Moreover, Dye (1988) suggests two reasons for top management to smooth earnings. First, management is expected to meet the external demand of higher stock returns. As smoother earnings are believed to increase a firm’s share price, it can be used by management to present higher cash flows to shareholders. The second reason for earnings smoothing relates to the internal demand for optimal contracting. Top managers may be unable to communicate all relevant information to investors. By giving top management the discretion to adjust earnings figures, management is able to communicate this information. However, it also enables them to manage earnings.

According to Jung et al (2013), earnings smoothing practices are harmful to a firm’s actual value, and therefore to its stakeholders, in two ways. First, earnings smoothing is a time consuming activity and attracts managers’ attention from activities that are truly beneficial to the firm. Second, managers take a risk by using earnings smoothing. If earnings smoothing is detected, this may deteriorate the value of the firm because its management loses credibility among stakeholders and rating agencies. Also, the firm may be fined when earnings smoothing is detected.

2.3.6 Big bath accounting

Big bath accounting may occur in years when a firm’s performance is very poor. Although it is illegal, a firm’s top management may then decide to ‘write off’ that financial year in the books. In that case, the practice of big bath accounting concerns making adjustments to earnings to deteriorate performance even further to allow for better earnings in subsequent years, as managers are suggested to believe that one very poor financial report is less harmful than a chain of mediocre reports (Arya et al, 1998). This allows to incur fewer costs in subsequent years. Kirschenheiter and Melumad (2001) describe how big bath accounting is applied for very bad news reports (i.e. large negative earnings surprise). In their paper, they develop a model for reporting strategies and analyze whether equilibrium reporting strategies exist. Based on the model, they find that both earnings smoothing and big bath accounting are equilibrium financial reporting strategies, and therefore both can be beneficial to a firm’s top management.

(27)

27

Big bath accounting is often accommodated in accruals due to their low degree of verifiability (Arya et al, 1998). Therefore, big bath accounting does not result from a cash transaction. For instance, big bath accounting may lead to excessively writing off on inventory and reversing this write off in subsequent years, or anticipating the economic effects of a future plant closing by accounting for it way in advance.

2.4 Expectations & hypotheses

In this section I summarize the previously discussed theories to form expectations for this study. Subsequently, I formulate the expectations into hypotheses.

2.4.1 Expectations

Taking into account the theories that are discussed in this study, several expectations can be formed. Based on the earnings smoothing and big bath accounting literature, the expectations related to the reporting for natural disasters are the following.

First of all, under the damage reporting of natural disasters in their financial statements, organizations are expected to apply earnings smoothing practices. In Chapter 2.1.2 I illustrated the degree of reporting flexibility that a natural disaster situation grants management, which according to Trueman and Titman (1988) is required to smooth earnings. This results in several manipulations. In case of bad news earnings announcements (i.e. negative earnings surprise), earnings is over reported by managers as part of earnings smoothing strategies (Trueman & Titman, 1988; Kirschenheiter & Melumad, 2001). Therefore, the damage due to the natural disaster is understated and is partly disclosed in subsequent years in order not to deviate excessively from the expected earnings. For good news reports, managers tend to under report earnings initially. Accordingly, the damage due to the natural disaster is overstated and is partly disclosed in subsequent years to enable meeting future earnings forecasts (Trueman & Titman, 1988; Kirschenheiter & Melumad, 2001). As good news turns into very good news, the under reporting is magnified to level with the expected earnings (Jung et al, 2013; Trueman & Titman, 1988; Kirschenheiter & Melumad, 2001).

(28)

28

Second, under the damage reporting of natural disasters in their financial statements, organizations are expected to apply big bath accounting practices. Accordingly, organizations ‘take a bath’ in case their announced earnings figures are far below expected earnings (Kirschenheiter & Melumad, 2001). This implies that they overstate the damage of the natural disaster to allow for higher profits in subsequent years (Arya et al, 1998). In subsequent years, part of the damage of the initially disclosed damage is reversed (Arya et al, 1998). In accordance with the foregoing, the expectations are stated in Table 1. Please note that I define the different news categories (bad news, good news, etc.) later on, in Chapter 3.3.1.

To provide a better understanding, the expectations are graphically displayed in Figure 1. This gives a very broad image of the expectations and should not be interpreted as an actual prediction. Additionally, please note that the expectations for the reporting of good news and bad news are not necessarily expected to be symmetric.

Contrary to these expectations, there are situations that incentivize managers to release news quicker in case of bad news. For example, the presence of high litigation risk may induce managers to immediately release all the bad news, as they are at risk when they withhold the bad news (Skinner, 2004; Healy & Palepu, 2001). However, the circumstance of litigation risk is not in the scope of this study, as this is not part of the research objective and because the presence of litigation risk is hard to measure in the case of reporting for natural disasters. Therefore, litigation risk possibilities are ignored for the purpose of this study.

(29)

29

TABLE 1

Expectations for initial under or over reporting of the damage of natural disasters

#

Earnings announcement perception

(Earnings surprise) Expectation for disaster year Supporting literature

1 Very good news

(large positive earnings surprise)

Large earnings smoothing (excessively overstate damage)

 Jung et al (2013)

 Kirschenheiter and Melumad

(2001)

 Trueman and Titman (1988)

2 Good news

(positive earnings surprise)

Small earnings smoothing (overstate damage)

 Kothari et al (2008)

(2001)

3 Neutral (industry

average;

no significant earnings surprise)

Fair reporting  Kirschenheiter and Melumad

(2001)

4 Bad news

(negative earnings surprise)

Earnings smoothing (understate damage)

 Kothari et al (2008)

(2001)

5 Very bad news

(large negative earnings surprise)

Big bath accounting (overstate damage)

 Arya et al (1998)

(2001)

Notes: The table gives the five different expectations of this study. For firms that are about to announce a (very) good news earnings report, I expect to observe (large) downward earnings smoothing in the concurrent year, as these firms have already (significantly) outperformed their targets. Therefore, the damage of the disaster is expected to be overstated (excessively). The firms that are about to release neutral news are very close to their targets and therefore are not expected to manipulate the damage of the disaster in order to achieve better results. The bad news firms, contrary to the good news firms, have not reached their targets. Therefore, these firms are expected to understate the damage of the disaster in order to move closer to their targets. For the very bad news firms, an inverse relationship is expected. These firms are very far away from reaching their targets and are expected to overstate the damage of the disaster in order to be able to report gains from the disaster in subsequent years.

(30)

30

Over report damage

Earnings smoothing

Big bath accounting

Bad news Good news (compared to industry)

Under report damage

FIGURE 1

Graphically displayed expectations for under or over reporting of the damage of natural disasters

Notes: This figure serves to provide a very broad image of the expectations and should not be interpreted as an actual prediction. The table displays the expected relationship between the earnings announcement news category and the under or over reporting of the damage of a natural disaster in the initial year. Please note that the expectations for the reporting of good news and bad news are not necessarily expected to be symmetric, as the figure may suggest.

(31)

31

2.4.2 Hypotheses

In the previous section, the expectations of this study are stated. For each earnings announcement category, the earnings management prediction is provided along with the direction of the under or over reporting of the damage. In this section, I formulate the expectations as viable hypotheses.

First of all, expectations are given for each good news and bad news earnings announcement scenario. The first hypothesis, which summarizes expectations 1-4, is as follows.

H1: Firms use earnings smoothing under reporting for natural disasters.

Specifying H1 enables to find more detailed results for earnings smoothing practices. For example, Kothari et al (2008) only acknowledge earnings smoothing in case of bad news. By breaking apart H1 into the three different implications of earnings smoothing, the (partial) results are stated in a more nuanced manner. Therefore, I divide H1 into the following three sub hypotheses:

H1a: In case of bad news, firms under report the damage of natural disasters initially. H1b: In case of good news, firms over report the damage of natural disasters initially. H1c: In case of very good news, firms over report the damage of natural disasters initially

even more than for good news.

Please note that hypothesis H1c is conditional on the findings related to hypothesis H1b. If it turns out that the obtained results are contrary to hypothesis H1b, then investigating H1c is of no use. The second hypothesis concerns expectation 5 and is as follows.

H2: In case of very bad news, firms apply big bath accounting and over report the damage initially.

For the support of the predicted directions of the hypotheses, please refer to the previous sections. An explanation of the research methodology and definitions of the variables is given in the next chapter.

(32)

32

3. Research methodology

In this section, I discuss the research methodology of this study. First of all, I describe the sample selection process, followed by a description of the sample. Subsequently, I come up with a basic regression model that allows for further elaboration on the variables. Then, I define the independent and dependent variable and the (explanatory) control variables that are part of the regression model. I conclude this section by explaining the detailed regression model.

3.1 Sample selection

For the purpose of this study, I select 100 large firms from different capital markets in the US that were possibly hit by Hurricane Katrina (from now on referred to as ‘the disaster’). The reason for this low sample size is that most of the data needs to be manually collected and therefore is very time-consuming. Firms are selected from different industries to create a diversified sample that can eventually justify generalization. To meet the requirements of the sample selection, firms should publish their financial statements online and report under US GAAP. This allows for easy access to the required data and for fair comparison.

Through different news sources I identified four industries that have been hit particularly hard by the disaster and therefore have a high availability of data. From these four industries, I select an equal amount of firms for the sample. First of all, The Economist (2005) reports that “The insurance industry is hit so hard, that insurance premiums are expected to rise worldwide”. The total loss in the insurance sector due to Hurricane Katrina is estimated to be approximately USD 40-60 billion (The Economist, 2005). Moreover, Bloomberg (2005) reports that several oil corporations have suffered heavily from the disaster due to the shutdown of production operations. Also, as about 90% of the local utility networks are destroyed, the regional utilities industry is among the heavily affected industries (International Risk Governance Council, 2009). Finally, Stutz (2005) reports that many casinos in the Gulf coast were destroyed or heavily damaged by the disaster, thereby leading to an estimated loss of 17.000 jobs. In addition, many hotels were closed due to the disaster. Because the gambling and hotel industry are largely intertwined I combine these two.

(33)

33

In this paragraph I identified four industries that suffered heavily from the disaster. Recapitulating, these are the insurance, oil, utilities and the hotel and gambling industry. Using Compustat, I randomly select firms from these industries to create a sample. I describe this sample in section 3.2.

3.2 Sample description

To start with, I generated a sample of 12,241 US companies from Compustat that published their financial statements in the period July, 2005 to June, 2006. First, I eliminated all non-US firms and firms for which no statutory seat is given by Compustat. Second, I filtered the sample by eliminating all firms with revenues lower than USD 50 million. I did this because larger firms are more likely to publish their financial statements online, due to their generally large investor community. Additionally, large firms are more likely to have been hit by the disaster because larger firms generally tend to operate in more regions in the United States. Third, by using SIC codes, I filtered firms per industry, consistent with the industries mentioned in the foregoing paragraph. This leads to the remaining population, as displayed in Table 2.

TABLE 2

Description of Compustat population for selected industries

Industry SIC codes Amount of firms (entire US / states LA, FL, TX, AL, MS)

Oil 1300-1399, 2911 140 /10

Utilities 4900-4999 289 / 25

Insurance 6300-6499 165 / 7

Hotel & Gambling 7000-7099, 7990 43 / 0

Notes: This table provides an overview of the Compustat firm population in the selected industries. The second column displays the filtered SIC codes. In the third column, the population size is stated for the entire US and for the states that were impacted by Hurricane Katrina (Louisiana, Florida, Texas, Alabama and Mississippi).

(34)

34

As draws from table 2, the size of the Compustat population is too small to select 100 firms just from the states that were hit by the disaster. Therefore, I only select local firms for the utilities industry. For the other industries, I select firms from the entire US. By using the open-source tool ‘Research Randomizer’ for each industry sample, I created four equal sized subsamples. These are stated in Table 3.

TABLE 3

Overview of selected firms per industry

Industry Amount of firms Local / US wide sample

Oil 25 US wide

Utilities 25 Local

Insurance 25 US wide

Hotel & Gambling 25 US wide

Notes: This table provides an overview of the selected firms for the sample of this study. For each of the four industries, 25 firms are selected. For the utilities industry, the Compustat database contains sufficient data to select local firms. For the other industries, a national (US) sample is selected.

For a complete overview of the sample of 100 firms, please refer to Appendix A. The sample selection process is summarized in Table 4.

(35)

35

TABLE 4

Summary of sample selection process

Description # Firms

‘Raw’ data set from Compustat 12,241

Elimination of non-US firms and firms for which no US state is given

(2,772) Elimination of firms with revenues < USD 50 mln, and firms for

which no revenue is given (4,244)

Elimination of firms in outscoped industries and duplicates (4,588)

Full sample of firms, divided into four industries 637

Sample after selection 100

Notes: This table summarizes the sampling process. The table shows the elimination of firms in the population to come from 12,241 firms to 100 eventually, which is the selected sample size. The number of 12,241 firms is the total of firms posted in the Compustat file for 2005.

3.3 Basic regression model

Consistent with the foregoing, I identify two variables in the causal relationship. The independent variable is the news category of the earnings announcement (NEWS). This can be very bad, bad, neutral, good or very good news. In the regression model, this independent variable NEWS influences the dependent variable Revision, which is the fraction of the initially reported damage of the disaster that a firm disclosed through revisions in subsequent years. In a basic form, this leads to the following model:

Revision = a0 + a1NEWS + Controls + ε , (1)

I provide an explanation and definitions for these variables in Chapter 3.3.1 (NEWS) and 3.3.2 (Revision). Initially, to purely test the interaction between these two variables I use the basic regression model without control variables. Afterwards, I identify control variables to explain the relationship between the independent and dependent variables. Therefore, I refer to the control

(36)

36

variables as ‘explanatory variables’. Eventually, I come up with a detailed regression model that includes the explanatory variables.

3.3.1 Independent variable – earnings announcement news category (NEWS)

The independent variable in the basic regression model is the variable NEWS. The question whether firms report good news or bad news depends on the level of the earnings surprise that firms’ earnings announcements generated compared to other sampled firms. Consistent with the previous chapter, positive earnings surprises are associated with good news; negative earnings surprises with bad news. In line with McQueen et al (2002) I use quintiles to categorize the firms’ earnings announcements into different classifications. These are the following:

Very bad news (VBAD): 0 < ESₓt < 20

Bad news (BAD): 20 < ESxt < 40

Neutral news (NEUT): 40 < ESₓt < 60

Good news (GOOD): 60 < ESxt < 80

Very good news (VGOOD): 80 < ESxt < 100

In these quintiles, ESxt represents the earnings surprise percentile of firm x compared to the other

sampled firms in year t. The larger the earnings surprise as compared to the other sampled firms, the better the classification of the earnings announcement (i.e. good news, bad news, etc.). The earnings surprise function can be defined as follows:

ESxt = (FExt – AExt) / AExt , (2)

where FExt represents forecasted earnings (a proxy for forecasted earnings is given later on) for

firm x in year t and AExt represents actual earnings. This function is consistent with Brown (1997).

To obtain the figures for the forecasted earnings, prior literature mentions two different approaches. The random walk model (Kilian & Taylor, 2003; Finn, 1986) gives an earnings forecast based on earlier earnings figures by applying growth factors to time-series data. The second approach is to observe analysts’ forecasts by obtaining the data from the I/B/E/S consensus file (Bradshaw et al, 2012; Field et al, 2005). This file contains forecasts from different prominent

Agency problems and the reporting for natural disasters : damage estimates of hurricane Katrina

Agency problems and the reporting for natural disasters:

Damage estimates of Hurricane Katrina

Contents