Abnormal stock returns and hurricanes event study of the five most destructive hurricanes to make landfall in the United States since 2005

This paper tests how stock prices respond to information that may affect a company's value in a material fashion. I evaluate stock price reactions to the five most destructive hurricanes to make landfall in the United States since 2005. The event study aims to analyze short-term stock price reactions of firms that are likely to benefit economically before and after the landfall of the hurricanes. I find the cumulative average abnormal return (CAAR) is statistically significant at a confidence level of 95%. Cumulative abnormal returns are generally realized following the first trading day during or after landfall. The average abnormal return for stocks on the day of landfall is also positive and significant.

ABNORMAL STOCK RETURNS AND HURRICANES EVENT STUDY OF THE FIVE MOST DESTRUCTIVE HURRICANES TO MAKE LANDFALL IN THE UNITED STATES SINCE 2005 by Jay T Anderson A Senior Honors Thesis Submitted to the Faculty of The University of Utah In Partial Fulfillment of the Requirements for the Honors Degree in Bachelor of Science In Finance Approved: ______________________________ Elizabeth Tashjian Thesis Faculty Supervisor _____________________________ Michael Cooper Chair, Department of Finance _______________________________ Elena Asparouhova Honors Faculty Advisor _____________________________ Sylvia D. Torti, PhD Dean, Honors College May 2017 ABSTRACT This paper tests how stock prices respond to information that may affect a company’s value in a material fashion. I evaluate stock price reactions to the five most destructive hurricanes to make landfall in the United States since 2005. The event study aims to analyze short-term stock price reactions of firms that are likely to benefit economically before and after the landfall of the hurricanes. I find the cumulative average abnormal return (CAAR) is statistically significant at a confidence level of 95%. Cumulative abnormal returns are generally realized following the first trading day during or after landfall. The average abnormal return for stocks on the day of landfall is also positive and significant. TABLE OF CONTENTS ABSTRACT ii INTRODUCTION 3 HURRICANES 6 COMPANIES 12 METHODOLOGY 15 RESULTS 24 CONCLUSION 35 EXHIBITS 36 REFERENCES 44 2 1. INTRODUCTION Event studies are an empirical tool used to assess the impact of an event on the value of a company. It is generally assumed that stock markets are efficient in reflecting the information about a company. The efficient market hypothesis states that at any time security prices reflect all available information on a company (Fama E. F., 1970). Central to the hypothesis is the conjecture that investors seek to maximize risk-adjusted returns. In theory, traders immediately revise the prices which they are willing to trade at, so stock prices adjust immediately. Is the landfall of one the five most destructive hurricanes to make landfall in the United States since 2005 material enough to affect the value of companies involved in repairing, preventing damage, or supplying restoration materials to electrical, residential, commercial and water infrastructure exhibit short-term abnormal returns around the landfall of a hurricane? I search for the link between information, the landfall of a hurricane, and company value. The null hypothesis states this link is not material. In my tests, I assume that risk is reflected in the stock price of a company through the market model, which controls for broad stock market movements and the sensitivity of a stock to those movements. I study five events for 13 to 17 companies across the five most destructive hurricanes by economic value since 2005. The event period tested is from fifteen days before hurricane landfall to five days after. I compare actual stock returns to expected returns as a result 3 of changes to the stock market as a whole. This analysis determines the effect that each hurricane has on the stock returns of companies involved in repairing, preventing damage, or supplying restoration materials to electrical, residential, commercial and water infrastructure, and tests whether the aggregate returns across the event horizon are statistically significant. I analyze the timing of abnormal returns to explain if and when news of the landfall of a hurricane affects the value of companies in my study. My results reject the null hypothesis, showing the information of the value of a hurricane affects the value of stocks. I conclude this by analyzing cumulative average abnormal returns (CAAR), which are positive and significant immediately following the landfall of the hurricanes for the pool of stocks I selected. Furthermore, average abnormal returns (AAR) are the largest and significant on the first day of trading on or after landfall. CAAR and AAR were used by Fama, Fisher, Jensen, and Roll in 1969, exploring how stock prices adjust to new information (Fama, Fisher, Jensen, & Roll, 1969). CAAR is a useful metric for understanding the impact of an event over the entire event window, while AAR shows abnormal returns during each day of the event. Inspiration for my thesis comes from research on short-selling activity around hurricanes Katrina and Rita (Blau, Benjamin M.; Van Ness, Robert A.; Wade, Chip, 2008). Shortselling is the process of borrowing shares of a company from an owner at the current market price and promising to replace the shares by subsequently buying them. This is a bet on the value of a stock going down. Blau, Van Ness, and Wade find abnormal short 4 selling of liability insurance two days after Hurricane Katrina and a substantial increase in short-selling before the landfall of Rita. An abnormal return is defined as the difference between the return an investor expects given the market’s performance and the stock’s risk and the actual return. Their research suggests insurance stocks decline in response to the news of hurricane landfall because traders respond quickly to the information, revising the prices at which they are willing to buy and sell. Other event studies have explored stock price reactions to hurricanes and various natural disasters. Reinhold P. Lamb’s work punctuates Blau, Van Ness, and Wade’s research on short-selling of property-liability insurer stock values around Hurricane Andrew. He concludes that property-liability insurers with direct policies written in Florida or Louisiana realized significant negative stock price reactions, while similar companies that were not exposed did not see negative jumps in stock price (Lamb, 1995). Hurricane Andrew devastated these two states, causing more than $20 billion in property damage (1992: A Year of Catastrophes, 1993). Further research has examined natural disasters and their impact on market returns, including Natural disasters – Blessings in disguise?, which shows that construction and materials industries’ earnings are positively affected by natural disasters, while travel and nonlife insurance industries are negatively affected (Koerniadi, Krishnamurti, & Tourani-Rad, 2012). Section 2 defines hurricanes and explains their impact on the economy. Section 2 also describes the time horizon, destruction, and financial impact of hurricanes in this study and presents a case study of hurricane Sandy and its impact on New Jersey’s economy. 5 Section 3 details the companies selected for this study. The event study methodology is discussed in Section 4, and results of the study are presented in Section 5. Section 6 provides a conclusion of findings and recommends actions for investors and suggests future research. 2. WHAT ARE HURRICANES? Tropical cyclones are enormous, swirling storms that form over warm ocean waters. These storms are called “hurricanes” in the North Atlantic. A storm is classified as a “tropical storm” when its sustained surface winds reach 39 mph. When sustained winds reach 74 mph, tropical storms attain the status of “hurricane.” Hurricane season starts on June 1 and ends on November 30, although hurricanes have been observed outside this time horizon (NOAA, 2016). The most important ingredient for generating a hurricane is sea-surface water temperature of at least 26.5 degrees C or 80 degrees F (Nordhaus, 2006). The central low pressure of a hurricane, the lowest barometric pressure, pulls warm and moist ocean air which causes thunderstorms to swirl around the center of the storm (Ocean Tides, n.d.). Barometric pressure is the pressure exerted by the atmosphere measured by a barometer. Generally, a lower central pressure leads to stronger wind speeds around the eyewall of a hurricane, which is the most destructive area of the storm surrounding the center of a hurricane. Hurricanes have made a name for themselves so far in the 2000s. 2005 produced a record total of hurricanes and storm damage in the United States (Nordhaus, 2006). Hurricanes 6 have increased in intensity, frequency and duration in the North Atlantic since the early 1980s (U.S. Global Change Research Program, n.d.). However, the link between hurricanes and ocean temperatures is very complex, and the connection between hurricanes and warming oceans is debated by researchers. As the earth’s temperature has risen over the last century, nearly all studies project hurricanes will have greater rainfall (U.S. Global Change Research Program, n.d.). Eight of the top 10 most inflation-adjusted costly hurricanes have occurred since 2004 (NOAA). Leading the way, Hurricane Katrina resulted in $151.6 billion in inflationadjusted damages in 2005 (Tribune Media Wire, 2016). Damage from hurricanes is primarily due to wind, storm surge, and rain. Hurricanes are categorized on the SaffirSimpson Hurricane Wind Scale from 1 to 5 based on the speed of their wind. Category 1 hurricanes (74 – 95 mph) can snap large trees, damage power lines, and damage roofs, while Category 4 or higher hurricanes (157 mph or higher) usually completely destroy framed homes, cause extended power outages, and leave communities uninhabitable for weeks or months (NOAA, n.d.). Strong hurricanes are forming at a more frequent rate. In 2015, a record of 22 hurricanes strengthened to Category 4 or 5 in the Northern Hemisphere (Dolce, 2015). The old record of 18 hurricanes was set in 2004. Along coasts, storm surge can be the greatest threat from a hurricane. Storm surge is an abnormal rise in water levels created by a hurricane, which can erode beaches and coastal highways, destroy marinas, and flood neighborhoods. In 2004, storm surge from Hurricane Katrina killed at least 1,500 people along the Louisiana coast, showing the 7 direct and indirect catastrophic potential of this phenomenon. Rainfall further adds to flooding caused by storm surge, but it can affect areas even further inland (NOAA, n.d.). The large amount of water dumped on land during a hurricane produces infrastructure damage, transportation disruption, and damage to anything it contacts. To provide perspective, Hurricane Matthew, a hurricane that made landfall in the United States in South Carolina in 2016, produced enough rain to fill the historical Rose Bowl football stadium 163,000 times (Rice, 2016). 2.1 IMPACT OF HURRICANES ON THE ECONOMY Clearly, hurricanes are devastating forces to structures. They can hinder a local economy for a lengthy time, displacing workers, destroying property and resources, and limiting transportation. This paper does not focus on industries hurt by hurricanes. It discusses sectors which aid in the prevention of destruction, survival, and rebuilding that occurs after a storm. A New York Times article from 2008, Do natural disasters stimulate economic growth?, explains that economies benefit in some ways after a natural disaster strikes (Bennett, 2008). In the article, Mark Skidmore, an economics professor at Michigan State University says, “When something is destroyed you don’t necessarily rebuild the same thing that you had.” There are few papers in this area of research, but intuitively there are businesses that benefit economically from replacing aging and damaged infrastructure with newer and sturdier infrastructure. 8 2.2 HURRICANES IN THE STUDY Hurricanes chosen for this study, Katrina (2005), Wilma (2005), Ike (2008), Irene (2011), and Sandy (2012) are the five most economically destructive hurricanes, by billions of USD reported by the National Oceanic and Atmospheric Administration’s respective tropical cyclone reports, to affect the United States since 2005 (Exhibit 1). The NOAA estimates total damage by using insured and uninsured loss reports. Hurricane Katrina was catastrophic and the costliest hurricane in the history of the United States. It first made landfall as a Category 1 hurricane in Florida on August 25, 2005. The hurricane then strengthened to Category 5 over the Gulf of Mexico before making landfall in Louisiana on August 29, 2005. For this study, August 29 is assumed to be the time of landfall, because over 99% of the value of estimated damage occurred following this landfall of the hurricane. Hurricane Katrina is best known for its destructive flooding in New Orleans resulting from an unusually high storm surge, which left 80% of the city flooded and directly killed 1,500 people (CNN, 2016). Insurance companies and the National Flood Insurance Program paid out over $55 billion in claims. Following the hurricane, the Gulf Coast of the United States received significant funding from FEMA to rebuild damaged infrastructure, including $10.2 billion given to the City of New Orleans (FEMA). Hurricane Wilma ravaged the United States shortly after Katrina in 2005, making landfall near Cap Romano Florida on October 24, 2005. This storm had the lowest central 9 pressure of any hurricane recorded in the Atlantic Basin. The northeastern Yucatan Peninsula in Mexico was devastated by this hurricane as well, but reliable information on damage is not available. Although only five deaths are directly attributed to the storm, Wilma ranks as the fifth most costly hurricane ever and the fourth most costly since 2005, as 98 per cent of South Florida lost electrical service (Pasch, Richard, Blake, Cobb III, & Roberts, 2006). Interruption of access to electricity directly affects commercial and industrial customers. Costs of interruption include labor and material costs and lost revenues incurred during an outage and emergency response and public health and safety activities associated with outages (LaCommare & Eto, 2004). Hurricane Ike peaked at Category 4 intensity, devastating islands in the southeastern Bahamas and Cuba before making landfall near Galveston Island, Texas as a Category 2 hurricane on September 13, 2008. The storm destroyed 10 offshore oil rigs and high water levels from storm surge affected nearly the entire Gulf Coast, and areas of the Texas coast were inundated by over 10 feet of water. This hurricane was famous for its rain, dumping over 18 inches just north of Houston. Ike is the fifth costliest hurricane to affect the United States since 2005, and 7.5 million households lost power across the storm’s wide reaching path post landfall in Texas, Ohio, Kentucky, Louisiana, Arkansas, Indiana and Pennsylvania (Ferrell, 2008). Before making landfall near Cape Lookout, NC on August 27, 2011, as a Category 1 hurricane, Hurricane Irene peaked as a Category 3 hurricane but weakened after crossing the Bahamas. Irene barreled up the eastern seaboard. Northeastern states up to New 10 England felt its effects, with catastrophic flooding in Massachusetts, New Jersey and Vermont. North Carolina was ravaged by the storm’s wind, rain and storm surge, and hundreds of millions of dollars in property was damaged in New York City and Long Island. The third costliest hurricane to affect the U.S. since 2005, Irene affected a very broad population, and nearly 6 million households lost power. A late-season storm, Hurricane Sandy traveled through the Caribbean as a Category 3 storm. Although it weakened to tropical storm level winds of approximately 80 mph before making landfall in Brigantine, NJ on October 29, it grew immensely while moving through the Bahamas and became the largest hurricane on record in the Atlantic Basin, reaching nearly 1,000 miles in diameter (Huffington Post, 2012). The storm’s incredible size caused devastating storm surge in New York and New Jersey, while all coastal states in the Northeast were affected by the surge. In the United States, over 650,000 homes were damaged or destroyed and 8.5 million customers lost power. Sandy was the second costliest hurricane since 2005 and sixth most costly hurricane ever to hit the United States after adjusting for inflation (NOAA). 2.3 CASE STUDY The publicly traded companies in this study are chosen from industries that are likely to benefit economically from a hurricane. In the short term, industries such as tourism, insurance, commercial trucking and manufacturing are negatively affected by a hurricane 11 (U.S. Department of Commerce, 2013). The following case study analyzes the economic impact Hurricane Sandy had in New Jersey. Sandy caused an estimated loss of $950 million in tourism spending in New Jersey in 2013, with reduced employment in accommodations, food services, retail, amusements and performing arts and transportation services sectors. Atlantic City, a major tourist destination in New Jersey, saw casinos and resorts close for many days, and the Jersey Shore required repair following destructive storm surge. New Jersey’s state government estimated that the state would spend $29.5 billion in construction costs to rebuild infrastructure. Spreading the construction costs over four years, this resulted in about 70,000 additional jobs per year. Furthermore, $5.5 billion in Federal aid was authorized for projects in New Jersey. The U.S. Department of Commerce did not find evidence of any long-term losses in the travel and tourism industries from the storm. The department found that although Sandy did not have long-term effects on industries, it did have shortterm economic effects on a wide variety of industries. Companies and homeowners were faced with repairing structural damage, draining flood water, waiting for electrical power restoration, and debris removal among other activities. 3. COMPANIES The null hypothesis of this study states hurricanes have no impact on the value of the firms in the study. I test companies involved in repairing, preventing damage, or supplying restoration materials to electrical, residential, commercial and water 12 infrastructure. After screening for companies whose business model fulfills this need, 17 securities are chosen that are conjectured to benefit economically from a hurricane. All securities in this study are publicly traded companies which trade on the New York Stock Exchange or the NASDAQ. I classify the chosen companies as a 17-company “hurricane index.” I use Bloomberg to screen for companies and search for news articles around the landfall of hurricane that mention companies in industries included in the study. The companies are primarily national firms that have business in areas which were affected by hurricanes. Inspiration for the “hurricane index” comes from comments by the CEO of Generac Holdings Inc. (NYSE: GNRC), the leading seller of generators in the home standby power market. During a conference call on October 31, 2012, the first trading day for the New York Stock Exchange following the landfall of Sandy, Generac’s chief executive officer, Aron Jagdfeld, said the hurricane would positively affect the company. He said sales for 2012 were now expected to rise “in the low 40 percent range” over the last year (Winter, 2012). Not only did sales for the company receive a boost prior to the hurricane, but sales were expected to rise following the storm as consumer’s awareness of the benefits of generators rose. Sandy caused 8.5 million households in the Eastern United States to lose power. The storm thwarted businesses’ ability to operate if they did not have standby generators. On October 31, 2012, Generac’s shares jumped 20.01%, compared to the S&P 500, which only returned 0.02%. Generac beat Jagdfeld’s initial expectations, and on February 14, 2013, the company announced that 2012 sales jumped 48.5% over 2011, largely driven by Sandy’s destruction (Generac Investor Relations, 2013). 13 Table 1 contains a list of companies used in this study and a brief description of their businesses. Table 1: Description of Companies in the Study Company Ticker Exchange Beacon Roofing Supply BECN NASDAQ Briggs & Stratton Corp. BGG NYSE Clean Harbors CLH NYSE EMCOR Group EME NYSE Flowserve Corporation FLS NYSE Generac Holdings GNRC NYSE Granite Construction GVA NYSE The Home Depot HD NYSE HD Supplying Holdings HDS NASDAQ Lowe's Companies LOW NYSE Business Description Residential and non-residential roofing material distributor Gasoline engine and standby power generators manufacturer and distributor Environmental cleanup services, emergency spill response, hazardous waste management Mechanical and electrical construction, industrial and energy infrastructure Movement, control and protection of flow of materials, maintenance of industrial pumps and vales Residential standby and industrial power generator manufacturer and distributor Heavy civil contractor and construction materials producer, large federal contractor Home improvement retailer Wastewater transmission, supplier of property repair products, hardware, tools and materials Home improvement retailer Engineering, fabrication, infrastructure, and maintenance services, electrical Matrix Service Co. MTRX NASDAQ infrastructure construction Manufacturer of glass fiber reinforcements Owens Corning OC NYSE for composites, insulation, and roofing Supplier and manufacturer of residential PGT Innovations PGTI NASDAQ impact-resistant windows and doors Manufacturer of water and fluid solutions, Pentair PNR NYSE valves and controls, and other equipment Provider of electrical infrastructure services, including design, installation, Quanta Services PWR NYSE upgrade and repair Designer, producer and marketer of Skyline Corporation SKY NYSE manufactured and modular housing Manufacturer of motor homes, travel Winnebago Industries WGO NYSE trailers, and commercial vehicles Manufacturer of water infrastructure, such as water pumps, treatment equipment, and Xylem XYL NYSE filtration *Business descriptions accessed from SEC filings and Yahoo Finance, February 25, 2017 14 4. METHODOLOGY The purpose of an event study is to analyze the impact events have on the value of companies. As explained by Brown and Warner, event studies provide a test of market efficiency (Brown & Warner, 1980). Thus, nonzero abnormal security returns which continue after an event are inconsistent with the hypothesis that security prices adjust quickly to fully reflect new information. If securities are priced efficiently, the magnitude of the impact of an event on a securities price should quantify the impact on the wealth of a firms’ shareholders. Abnormal returns capture the impact of an unanticipated event on firm value. Abnormal returns are defined as security returns which differ from expectations given a prediction model. For this study, I use the market model to predict “no-news” returns, which assumes a hurricane did not affect the value of the firms in my sample. The abnormal return of a firm i and event date t is the difference between the actual realized return and expected return in absence of the event (Event Study Metrics, n.d.). 𝐴𝑅𝑖,𝑡 = 𝑅𝑖,𝑡 − 𝐸(𝑅𝑖,𝑡 ) Where, ARi,t = the abnormal return for firm i at time t; Ri,t = the observed return for firm i at time t; and, E(Ri,t) = the expected return in absence of the event for firm i at time t. 15 Before defining the market model which computes expected returns for the model, I define the underlying theoretical model of security prices. The Capital Asset Pricing Model (CAPM) is derived based on a set of assumptions: capital markets are perfectly competitive, investors have homogeneous expectations given a set of circumstances, and security returns are normally distributed. The CAPM provides the return an investor expects based on the relationship between a security’s beta, the risk-free rate, and the equity risk premium. The framework was developed by William Sharpe (1964), Jack Treynor (1962), John Lintner (1965a, b) and Jan Mossin (1966) (Perold, 2004). 𝐸(𝑅𝑖 ) = 𝑅𝑓 + 𝛽𝑖 ∗ [𝐸(𝑅𝑚 ) − 𝑅𝑓 ] Where, E(Ri) = the return required on security i; Rf = the risk-free rate of return; βi = beta for security i; and, E(Rm) = the market return. The CAPM is an ex-ante model, because it refers to future expected returns for a company. The market model is an ex-post empirical version of the CAPM. The market model is a linear model in which returns are normal, and the best estimates for the intercept and slope of the model are the standard regression parameters. Essentially, the market model uses the parameters to “allow for” changes in firm i’s stock price relative to the stock market. The equation for the market model is below (Event Study Metrics, n.d.). 16 𝑟𝑖,𝑡 = 𝛼𝑖 + 𝛽𝑖 𝑟𝑚,𝑡 + 𝜀𝑖,𝑡 With 𝐸[𝜀𝑖,𝑡 ] = 0 and 2 𝑉𝐴𝑅[𝜀𝑖,𝑡 ] = 𝜎𝜀,𝑖 where, ri,t = the expected return unconditional on the event for firm i at time t; αi = y-intercept for security i; βi = slope coefficient for security i; rm,t = return on the index at time t; and, εi,t = the error term for security i at time t. Expected returns are based on the underlying return-generating model, conditioned on a particular information set. The information set is usually past asset returns. The return generating model I use is the market model. The calculation of the parameters of this model is explained further in the methodology section. It is important to understand the time periods of an event study. The estimation window L1 provides the basis for estimating the expected return generating process during the event window L2. Estimation and event windows are non-overlapping to ensure the benchmark return is independent of the computed abnormal return. Lack of independence 17 invalidates the standard statistical methods used to test whether the abnormal return is statistically different from zero. The estimation window is a period of trading days prior to the event date that is used to estimate the market model forecast given a day’s market return for each security. There is not an academic consensus regarding the length of the estimation window (Event Study Metrics, n.d.). For this paper, I use an estimation window of 60 days, 15 to 75 days before each event window. The estimation window for Hurricane Wilma is adjusted to omit the days 34 to 54 days before the event, because this is the event window for Hurricane Katrina. The event window used is from 15 days before the landfall of a hurricane to 5 days after landfall. This is used because it sufficiently captures the life span of each hurricane for the five events studied. None of the hurricanes formed as a tropical depression, which is the earliest stage of a tropical cyclone, earlier than 15 days prior to landfall, and all storms dissipated within 5 days of landfall. The post event window can be used to capture any additional abnormal returns resulting from the event, which can be caused by the release of additional information concerning an event. 18 The day of landfall, t = 0, is the first trading day during or after the landfall of a hurricane. It is important to note that days in the study are expressed as trading days when the New York Stock Exchange is open, Monday through Friday. Hurricane Sandy presents a unique case, because the stock market closed for two days on October 28 – 29, 2012, as the hurricane’s damage was catastrophic in the New York City area. This was the first time since 1888 that weather forced the NYSE to close for two days (Kim, 2012). October 30, 2012, is used as t = 0 for hurricane Sandy, the first day the NYSE was open following landfall. Historical stock data is acquired for target companies from Bloomberg. The valueweighted Center for Research in Security Prices (CRSP), representing nearly 100% of the investable U.S. equity market and commonly used in academic research, is used as a proxy for the market. Researchers and investors may decide to use different benchmarks such as the S&P 500 or Dow Jones Industrial Average, but I use the CRSP because it is the most comprehensive U.S. equity index available (University of St. Thomas, n.d.). Security returns are calculated as follows and do not account for dividends. Day 2 Closing Price – Day 1 Closing Price Day 2 Closing Price As previously mentioned, my study uses the market model to drive predictions for individual asset returns during each event window. This model assumes a constant linear relationship between individual security returns and the return of the CRSP. I assume the 19 dependent variable Y, the individual company’s return, is linearly related to the independent variable X, the CRSP return. To determine the parameters of the market model, a regression is computed between the stock price and CRSP for each day of the estimation window using the following equation. 𝑌𝑖,𝑡 = 𝛼𝑖 + 𝛽𝑥𝑡 + 𝜀𝑖,𝑡 Where, Yi,t = the dependent variable (security i return from hurricane index) at time t xt = the independent variable (return on CRSP index) at time t αi = y-intercept for security i β = the slope of the line εi,t = the error term; E(εi,t) = 0 The return on the CRSP is the independent variable and the return on each security is the dependent variable, αi and βi are constant for each trading day, and εi,t is how far above or below the line the security’s return is each day. Under the assumptions for an Ordinary Least Squares regression, the errors in the regression have a conditional mean of zero. β and α are the coefficients that minimize the sum of squared errors in the data. Below is an example of a linear regression calculated for the return of Home Depot and the CRSP during the estimation period for Hurricane Katrina. The line is a statistical line that minimizes squared errors during the estimation period of the event study. 20 Table 2: Home Depot Example Regression Home Depot Daily Return HD Regression -1.5% y = .0002 + 1.6845x 5.0% 4.0% 3.0% 2.0% 1.0% -1.0% -0.5% 0.0% 0.0% -1.0% 0.5% 1.0% 1.5% -2.0% -3.0% CRSP Daily Return I run each regression independently. 17 companies and 5 events were selected for this study but I only run 74 regressions. Some companies were not publicly traded at the time of the event. Table 3 contains a list of the securities used for each hurricane. 21 Table 3: Securities Traded by Hurricane Ticker BECN BGG CLH EME FLS GNRC GVA HD LOW MTRX OC PGTI PNR PWR SKY WGO XYL Total Katrina x x x x x x x x x x x x x 13 Sandy x x x x x x x x x x x x x x x x x Ike x x x x x Wilma x x x x x x x x x x x x x x x x x x x x x x x Irene x x x x x x x x x x x x x x x x 17 15 13 16 Table 4 is a sample of the regressions from Hurricane Katrina’s estimation window. The dependent variable is the individual security return, the independent variable is the CRSP return, the slope is β, the intercept is α, standard errors measure deviations from predictions, and R-squared explains the amount of variation in Y explained by x. 22 Table 4: Hurricane Katrina Example Regression Ticker BECN BGG CLH EME FLS GVA HD LOW MTRX PNR PWR SKY WGO Y-intercept -0.001 -0.001 0.001 0.000 0.001 0.004 0.000 0.001 0.006 -0.001 0.002 -0.001 0.001 Beta 2.451 1.301 2.052 1.906 1.551 2.189 1.685 1.614 0.252 0.723 1.703 1.847 1.811 Standard error 0.026 0.011 0.023 0.010 0.011 0.016 0.010 0.011 0.028 0.018 0.019 0.015 0.020 R-squared 0.223 0.308 0.207 0.553 0.390 0.397 0.504 0.440 0.003 0.050 0.216 0.348 0.217 A higher R-squared indicates how well the returns of a stock are explained by the CRSP returns. Thus, an R-squared of .50 means that 50 percent of a security’s movements are explained by the index. The R-squared for some of the securities are very low, such as 0.05. This is not uncommon, because security prices can be very volatile and are not always explained by the index. The standard error is sometimes easier to interpret than the R-squared, and it represents the variability in Y that unexplainable (Frost, 2014). The model explains a lot of variation if standard error is low. The market model, explained earlier in the methodology section, is then used to calculate the expected return of each security during the event window for each hurricane by using the parameters calculated through the Ordinary Least Squares regression. The market model takes into account both market wide factors and the systematic risk of each 23 security and is driven by the slope β and the intercept α, which were calculated during the estimation window (Brown & Warner, 1980). The equation for the market model is listed below again. 𝑟𝑖,𝑡 = 𝛼𝑖 + 𝛽𝑖 𝑟𝑚,𝑡 + 𝜀𝑖,𝑡 With 𝐸[𝜀𝑖,𝑡 ] = 0 and 2 𝑉𝐴𝑅[𝜀𝑖,𝑡 ] = 𝜎𝜀,𝑖 where, ri,t = the expected return unconditional on the event for firm i at time t; αi = the y-intercept for security i; βi = the slope coefficient for security i; rm,t = the return on the index at time t; and, εi,t = the error term for security i at time t. 5. RESULTS Abnormal returns are computed for the event window, from 15 days prior to landfall to 5 days after landfall. I calculate the expected return using the CRSP return and α and β parameters estimates from the estimation window. Then, I subtract these returns from the actual returns for each stock and each day. I compute the daily average abnormal returns 24 for each company during the event window. This is the same methodology used in the study Hurricanes as News? A Comparison of the Impact of Hurricanes on Stock Returns of Energy Companies (Liu, Ferreira, & Karali, 2015). This metric is useful, because it allows an analysis of abnormal returns for each day of the event window, particularly on the day of landfall, t = 0. The formula used is below. 𝑛 1 𝐴𝐴𝑅𝑡 = ∑ 𝐴𝑅𝑖,𝑡 𝑛 𝑖=1 Where, AARt = the average abnormal return for company i at time t; ARit = the abnormal return for company i at time t; and, n = the sample size. To understand the cumulative impact of the hurricanes during the event window, we compute the firm-specific cumulative abnormal returns (CARs). CAR is defined by the formula below. 𝐶𝐴𝑅𝑖,𝑡 = 𝐶𝐴𝑅𝑖,𝑡−1 + 𝐴𝑅𝑖,𝑡 Where, CARi,t = the cumulative abnormal returns for company i at time t; CARt-1 = the cumulative abnormal returns for company i at t-1; and, ARt = the abnormal return for company i at time t. 25 Next, I compute CAAR on a rolling basis for each day of the hurricane. I do this to better understand when, if any, cumulative abnormal returns are realized in the index relative to the landfall of the hurricane. CAAR is the cross-sectional average of cumulative average abnormal returns of companies across the hurricanes. CAAR is defined below. 𝐶𝐴𝐴𝑅𝐻(𝑡1,𝑡2) 𝑁 1 = ∑ 𝐶𝐴𝑅𝑖(𝑡1,𝑡2) 𝑁 𝑖=1 Where, CAARH(t1, t2) = the cross-sectional cumulative average return for the “hurricane index” H from time t1 to time t2; CAARi(t1,t2) = the cumulative average abnormal return for company i from time t1 to time t2; and, N = the number of securities. I compute CAAR and AAR under two conditions, because all stocks were not publicly traded during all five hurricanes. To test the robustness of the study, I compute AAR and CAAR for names only present for all five events and separately for all 17 companies. The AARs and CAARs for companies which were publicly traded for all five events are listed in Table 5. 26 Table 5: Average Abnormal Returns and Cumulative Average Abnormal Returns only for Securities Traded During all Hurricanes t -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 AAR -0.07% -0.51% -0.15% -0.19% 0.17% 0.13% 0.15% 0.12% 0.05% -0.20% -0.04% 0.22% 0.36% 0.13% 0.50% 1.46% -0.52% 0.93% -0.81% 0.14% 0.91% CAAR -0.07% -0.58% -0.73% -0.92% -0.75% -0.62% -0.47% -0.35% -0.29% -0.50% -0.54% -0.32% 0.05% 0.18% 0.68% 2.14% 1.62% 2.55% 1.74% 1.88% 2.78% *Excludes GNRC, PGTI, OC, XYL 27 Below are the AARs and rolling CAARs for securities across hurricanes for each day during the event window for all securities in the study. Table 6: Average Abnormal Returns and Cumulative Average Abnormal Returns for all Securities t -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 AAR -0.21% -0.56% -0.19% -0.03% 0.34% 0.12% 0.17% -0.03% 0.19% 0.02% -0.11% 0.47% 0.20% 0.28% 0.35% 2.46% -0.51% 0.53% 0.37% -0.01% 0.54% CAAR -0.21% -0.76% -0.95% -0.98% -0.63% -0.51% -0.33% -0.37% -0.18% -0.16% -0.27% 0.20% 0.40% 0.68% 1.03% 3.49% 2.98% 3.51% 3.88% 3.86% 4.41% *Includes all 17 securities A cross-sectional t-test is used to analyze whether the AARs and rolling CAARs in each period are significant. The t-test tests whether the difference between two variable’s averages reflects a difference in the population from which they were sampled. The computed t-statistic tells if the difference between the average is unlikely to have happened because of random chance. A large sample size, substantial difference between means, and a low standard deviation in average values enhance the likelihood of the 28 difference in average values being significant. The difference in means’ significance is tested at a 95% level of confidence. The significance of average abnormal returns is resolved for each day from t= -15 to t = 5. The formula below is used to calculate t-statistics for each average abnormal return (The University of Queensland). 𝑡𝐴𝑅 = 𝐴𝐴𝑅𝑡 𝜎𝐴𝑅 ⁄√𝑛 Where, tAR = the t-statistic; AARt = the average abnormal return for time t; σAR = the standard deviation of abnormal returns at time t; and, n = the sample size. I test the significance of overall cumulative abnormal returns on a rolling basis throughout the event window. The method is the same used by Robinson and BanwayoSkeete (2016) in their analysis of the financial impact of natural disasters on stock markets in the Caribbean (Robinson & Banwayo-Skeete, 2016). The formula for the t-test I use to calculate t-statistics for each rolling cumulative average abnormal return (CAARt) is below. 29 𝑡𝐶𝐴𝐴𝑅 = 𝐶𝐴𝐴𝑅(𝑡1,𝑡2) 𝜎𝐶𝐴𝐴𝑅(𝑡1,𝑡2) ⁄√𝑛 Where, tCAAR = the CAAR t-statistic; CAAR(t1,t2) = the cumulative average abnormal return from time t1 to t2; σCAAR = the standard deviation of cumulative average abnormal returns from time t1 to t2; and, n = the sample size. The null hypothesis assumes risk is reflected in the stock price of a company through the market model, which controls for broad stock market movements and the sensitivity of a stock to those movements, and assumes hurricanes have no impact on the value of the firms in the study After computing the t-statistics, I test the null hypothesis (H0). The null hypothesis states hurricanes have no effect on the value of the “hurricane index” and stock prices are driven by the market model. Thus, the alternative hypothesis (Ha) states the event of a hurricane has a material impact on firm value, which is reflected in abnormal returns of the “hurricane index.” Based on the t-statistics, a one-tailed t-test is calculated to test whether the abnormal returns were greater than zero. The t-statistic is compared to a critical value. Generally, in event studies, researchers use an alpha level of 0.05. This means that, if the null hypothesis is true and the statistical assumptions hold, the test statistic will fall below the critical value 95% of the time. Hence, I reject the null 30 hypothesis (H0) in favor of the alternative hypothesis (Ha) if CAAR and AAR are significant at a 95% confidence interval. Table 7: One Tailed t-Test (95% Confidence): Only Securities Traded During all Hurricanes t -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 AAR -0.07% -0.51% -0.15% -0.19% 0.17% 0.13% 0.15% 0.12% 0.05% -0.20% -0.04% 0.22% 0.36% 0.13% 0.50% 1.46% -0.52% 0.93% -0.81% 0.14% 0.91% t-Stat -0.35 -1.80 -0.83 -0.59 0.48 0.50 0.31 0.38 0.22 -1.17 -0.20 0.59 1.65 0.61 2.12 4.02 -1.61 2.24 -2.46 0.24 4.56 Significant? YES YES YES YES CAAR -0.07% -0.58% -0.73% -0.92% -0.75% -0.62% -0.47% -0.35% -0.29% -0.50% -0.54% -0.32% 0.05% 0.18% 0.68% 2.14% 1.62% 2.55% 1.74% 1.88% 2.78% *Excludes GNRC, PGTI, OC, XYL 31 t-Stat -0.35 -1.72 -1.82 -2.73 -1.47 -1.01 -0.81 -0.43 -0.31 -0.52 -0.52 -0.24 0.04 0.15 0.58 1.65 1.40 2.10 1.35 1.70 2.50 Significant? YES YES Table 8: One Tailed t-Test (95% Confidence): All Securities t -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 AAR -0.21% -0.56% -0.19% -0.03% 0.34% 0.12% 0.17% -0.03% 0.19% 0.02% -0.11% 0.47% 0.20% 0.28% 0.35% 2.46% -0.51% 0.53% 0.37% -0.01% 0.54% t-Stat -1.15 -1.71 -1.22 -0.08 1.03 0.62 0.44 -0.13 0.47 0.11 -0.58 1.51 0.67 1.14 0.79 2.19 -0.93 0.99 0.44 -0.03 1.63 Significant? YES CAAR -0.21% -0.76% -0.95% -0.98% -0.63% -0.51% -0.33% -0.37% -0.18% -0.16% -0.27% 0.20% 0.40% 0.68% 1.03% 3.49% 2.98% 3.51% 3.88% 3.86% 4.41% t-Stat -1.15 -2.20 -2.36 -2.51 -1.45 -1.00 -0.70 -0.60 -0.23 -0.20 -0.31 0.19 0.41 0.65 1.00 1.87 1.60 1.71 1.48 1.66 1.87 Significant? YES YES *Includes all 17 securities The t-tests show the cumulative average abnormal returns (CAARs) are significantly greater than zero at the end of the event window. Furthermore, the CAAR for all companies is also significant on the first trading day during or after landfall of a hurricane, but CAAR is significant two trading days after landfall excluding securities that are not traded during all five hurricanes. Also when t=0, results exhibit significant average abnormal returns (AARs) that are greater than zero. 32 The addition of Generac (GNRC) to the study in Table 8 could be a potential explanation for the discrepancy in results between the two tests. GNRC contributes significantly to the CAAR for the group of securities following the landfall of Sandy and Irene. During the first trading day following Irene (t=0), GNRC’s cumulative abnormal return (CAR) reaches 6.11% and following Sandy (t=0) it is 28.78%. The increased sample size from 13 to 17 companies is also a likely contributor to the difference, because a larger sample increases the power of the significance test. The power is increased because the larger sample size narrows the distribution of the test statistic. A higher power reduces the probability of a Type II error, which is an incorrect acceptance of a false null hypothesis (Calkins, 2005). In this case, although the point estimates of the CAAR is higher for the larger sample, the computed t-statistic is lower, reflecting a higher standard error in the larger sample. This is largely driven by the extremely high returns for Generac which increase the average return but also increase the standard error. Although the full sample is larger, 17 firms is still a relatively small number of firms and so adding one very volatile firm to the index reduces the precision of the point estimate. Although both Table 8 and Table 9 show significant abnormal average returns when t=0, the pool of 13 securities traded during all hurricanes also exhibits significant AARs when t=-1, 2, and 5. A possible explanation for this is the dilution of AARs by the addition of extra securities to the study which did not realize abnormal returns on those dates. To better understand the significant of AAR and CAAR around the landfall of the hurricanes and their robustness, t-tests are conducted from t = -2 to t = 2. 33 Table 9: One Tailed t-Test (95% Confidence) for t=-2 to t=2: Only Securities Traded During all Hurricanes t -2 -1 0 1 2 AAR 0.13% 0.50% 1.46% -0.52% 0.93% t-Stat 0.61 2.12 4.02 -1.61 2.24 Significant? YES YES YES CAAR 0.13% 0.63% 2.09% 1.57% 2.50% t-Stat 0.61 1.91 4.12 2.38 3.56 Significant? YES YES YES YES *Excludes GNRC, PGTI, OC, XYL Table 10: One Tailed t-Test (95% Confidence) for t=-2 to t=2: All Securities t -2 -1 0 1 2 AAR 0.28% 0.35% 2.46% -0.51% 0.53% t-Stat 1.14 0.79 2.19 -0.94 0.99 Significant? YES CAAR 0.28% 0.63% 3.09% 2.58% 3.11% t-Stat 1.14 1.19 2.07 1.53 1.59 Significant? YES *Includes all 17 securities These t-statistics reveal significant and positive CAAR and AAR during the first trading day following landfall for both groups of securities (t = 0). Therefore, while positive abnormal returns are seen in the days leading up to landfall, they are not significant until the day of trading following landfall. AAR and CAAR are generally higher for Table 10. The positive AARs and CAARs in Table 10 are likely less significant because of the increase in sample size from 13 to 17 companies. This introduces more noise to the data, increasing the standard deviation of returns and reducing the value of t-statistics. 34 6. CONCLUSION This research makes two contributions to the study of natural disasters and their impact on security performance: a hurricane’s anticipated impact on areas of an economy and recommendations for future research. First, this study presents an effort to understand the value effect hurricanes have on companies involved in assisting or providing means for disaster prevention and relief to businesses, homeowners, and governments. This study shows areas of the economy that are expected to outperform prior expectations following a hurricane. My findings are consistent with prior research, particularly the case study of Sandy’s impact on New Jersey’s economy by the U.S. Department of Commerce in 2013. Second, this research could be extended to test whether the market returns translate into real profits. This study could be to broadened to explore how powerful a hurricane must be to cause positive abnormal returns by testing less destructive hurricanes and hurricanes that did not meet meteorologist’s expectations. For example, Generac’s stock price increased 12% before the landfall of Hurricane Matthew in October of 2016 when the S&P 500 was flat. The storm’s track shifted and Florida was spared from its worst winds, and Generac’s stock dropped to a price close to its price before the hurricane formed (Cherwa, 2016). Finally, a study of intra-day trading data relative to the landfall of a hurricane, hurricane warnings issued by the NOAA, and news releases could shed more light on how the market assimilates the information into prices 35 7. EXHIBITS Exhibit 1: Hurricane Data Hurricane Katrina Wilma Ike Irene Sandy Dates August 23-30 , 2005 October 15-25 , 2005 September 1-14, 2008 August 21-28, 2011 October 22-29 , 2012 Damage (USD, Billions) Fatalities (Direct) Homes Lost Power (Millions) Landfall Category Category at Landfall August 25 (FL); August 29 (LA) 5 3 $108 1500 2.6 October 24 (FL) 5 3 $21.0 5 3.2 September 13 (TX) 4 2 $29.5 21 7.5 August 27 (NC) 3 1 $15.8 41 5.8 October 29 (NJ) 3 1 $50 72 8.5 *Source: NOAA 36 Exhibit 2: Rolling Cumulative Average Abnormal Returns (CAAR) by Company t -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 HD -0.26% 0.17% 0.11% -0.22% -0.13% 0.75% 2.88% 4.19% 4.91% 5.03% 5.82% 6.40% 6.57% 6.71% 5.75% 6.99% 5.10% 5.11% 4.75% 3.86% 4.33% LOW -0.75% -0.45% -0.15% -0.46% 0.11% 0.63% 2.43% 3.50% 3.80% 3.47% 3.96% 3.79% 3.28% 3.29% 2.48% 3.46% 1.67% 1.85% 1.44% 0.66% 1.40% BECN 0.55% -2.28% -1.72% -1.27% -2.16% -1.65% -1.76% -0.94% 0.25% 0.14% -0.11% 0.89% 2.43% 3.19% 3.43% 6.47% 5.69% 7.19% 5.38% 3.04% 5.40% SKY -0.52% -1.40% -3.28% -3.22% -5.15% -6.05% -3.51% -3.34% -3.61% -4.50% -4.78% -4.60% -3.83% -4.44% -2.93% -2.84% -3.31% -3.30% -3.89% -3.02% -1.65% WGO BGG 1.16% -0.49% 1.88% -0.01% 2.06% -0.25% 0.00% -2.97% -0.94% -1.43% -1.57% 0.40% 0.88% 0.95% 2.79% 1.51% 4.37% 2.03% 3.48% 2.91% 3.03% 3.57% 5.83% 5.88% 4.58% 5.44% 4.54% 5.82% 4.84% 7.99% 6.84% 10.12% 5.48% 8.82% 6.99% 10.84% 6.57% 10.43% 8.78% 7.99% 10.05% 9.21% 37 FLS 0.84% 0.37% 0.44% 1.03% 1.96% 0.55% 0.40% -0.19% -0.60% -0.57% -2.13% -3.56% -2.36% -2.48% -2.08% -2.60% -1.80% -2.40% -3.70% -1.17% -1.09% GVA 0.39% -1.06% -0.70% -1.62% -1.76% -1.96% -2.80% -2.74% -1.92% -3.10% -3.30% -2.65% -1.99% -2.25% -1.70% 0.07% 1.61% 3.78% 1.65% -0.20% 1.48% Exhibit 2 continued t -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 PNR -0.62% -0.48% -0.89% -0.92% -0.44% -0.48% -1.14% -1.46% -1.77% -1.24% -0.63% -0.60% -0.80% -0.79% 0.34% 1.08% -0.45% 2.48% 1.08% 1.31% 1.90% PWR -0.88% -1.83% -1.35% -0.96% -1.94% -2.63% -3.67% -5.17% -6.56% -6.27% -6.98% -8.56% -7.74% -7.36% -6.36% -3.58% -2.49% -2.43% -2.30% -2.74% -1.46% EME 0.23% 0.42% 0.05% 0.18% 0.16% -0.04% -0.24% -0.24% -1.05% -1.89% -2.87% -3.18% -2.30% -0.15% 0.48% 0.56% 1.09% -0.27% 0.56% -0.95% -0.80% MTRX -0.97% -2.43% -3.15% -1.38% 1.59% 2.88% -0.87% -3.29% -4.00% -3.77% -2.70% -4.27% -3.35% -4.53% -4.49% -3.84% -5.16% -1.44% -4.88% -0.14% 0.67% CLH 0.44% -0.42% -0.62% -0.10% 0.42% 1.11% 0.36% 0.89% 0.35% -0.16% 0.10% 0.49% 0.68% 0.75% 1.04% 5.05% 4.78% 4.72% 5.49% 6.99% 6.76% 38 GNRC PGTI -1.31% -1.26% -0.35% -0.76% -1.26% -0.14% -1.79% 2.57% 0.78% 0.89% 0.71% 1.56% 1.09% 0.00% 0.67% -0.63% -0.65% 4.90% 0.55% 5.89% -1.25% 6.12% 1.02% 6.74% 3.40% 5.19% 6.27% 5.51% 9.04% 0.10% 28.78% 1.28% 28.25% -5.63% 29.34% -11.82% 41.35% -7.23% 37.84% -3.70% 38.14% -7.67% OC -0.34% -4.44% -5.12% -4.21% -3.00% -2.85% -2.26% -1.76% -2.05% -1.31% -1.41% -0.13% 0.59% -0.47% 2.04% 1.23% 3.32% 3.30% 2.31% 0.84% 2.35% XYL 0.28% 0.12% -0.18% -1.25% 0.32% 0.00% 1.57% -0.06% -1.47% -1.40% -1.10% -0.11% -2.93% -2.04% -2.45% 0.27% 3.75% 5.76% 6.93% 6.31% 5.90% Exhibit 3: Average Abnormal Return (AAR) by Company t -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 HD -0.26% 0.43% -0.06% -0.33% 0.09% 0.88% 2.13% 1.31% 0.72% 0.12% 0.79% 0.58% 0.17% 0.14% -0.96% 1.25% -1.89% 0.01% -0.36% -0.89% 0.47% LOW -0.75% 0.30% 0.30% -0.30% 0.57% 0.52% 1.80% 1.07% 0.30% -0.33% 0.50% -0.17% -0.51% 0.01% -0.81% 0.98% -1.78% 0.17% -0.40% -0.78% 0.74% BECN 0.55% -2.83% 0.56% 0.44% -0.89% 0.51% -0.11% 0.82% 1.18% -0.10% -0.25% 1.00% 1.53% 0.76% 0.25% 3.04% -0.78% 1.50% -1.81% -2.33% 2.36% SKY -0.52% -0.88% -1.88% 0.06% -1.94% -0.89% 2.54% 0.17% -0.27% -0.89% -0.28% 0.19% 0.76% -0.61% 1.51% 0.10% -0.47% 0.01% -0.59% 0.87% 1.37% WGO 1.16% 0.72% 0.19% -2.07% -0.93% -0.63% 2.45% 1.92% 1.58% -0.89% -0.45% 2.80% -1.25% -0.04% 0.31% 2.00% -1.37% 1.51% -0.42% 2.21% 1.27% 39 BGG -0.49% 0.48% -0.24% -2.72% 1.54% 1.83% 0.54% 0.56% 0.52% 0.88% 0.66% 2.31% -0.45% 0.38% 2.17% 2.13% -1.29% 2.02% -0.41% -2.44% 1.22% FLS 0.84% -0.47% 0.08% 0.58% 0.94% -1.41% -0.15% -0.60% -0.41% 0.03% -1.56% -1.43% 1.19% -0.12% 0.40% -0.52% 0.80% -0.61% -1.30% 2.53% 0.08% GVA 0.39% -1.45% 0.37% -0.93% -0.14% -0.20% -0.84% 0.06% 0.82% -1.18% -0.20% 0.65% 0.66% -0.26% 0.55% 1.77% 1.54% 2.17% -2.13% -1.85% 1.68% Exhibit 3 continued t -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 PNR -0.62% 0.15% -0.41% -0.03% 0.48% -0.04% -0.66% -0.33% -0.31% 0.53% 0.61% 0.02% -0.19% 0.00% 1.13% 0.74% -1.53% 2.93% -1.40% 0.23% 0.59% PWR -0.88% -0.96% 0.48% 0.39% -0.99% -0.69% -1.04% -1.51% -1.39% 0.30% -0.71% -1.58% 0.82% 0.38% 1.00% 2.78% 1.09% 0.06% 0.13% -0.45% 1.29% EME 0.23% 0.20% -0.37% 0.13% -0.02% -0.20% -0.20% 0.00% -0.81% -0.84% -0.98% -0.31% 0.88% 2.15% 0.63% 0.08% 0.53% -1.35% 0.83% -1.51% 0.15% MTRX -0.97% -1.46% -0.72% 1.77% 2.97% 1.30% -3.75% -2.42% -0.71% 0.24% 1.07% -1.57% 0.92% -1.17% 0.03% 0.66% -1.33% 3.72% -3.44% 4.75% 0.81% CLH 0.44% -0.87% -0.20% 0.53% 0.51% 0.69% -0.75% 0.53% -0.54% -0.51% 0.26% 0.39% 0.19% 0.07% 0.29% 4.01% -0.27% -0.06% 0.77% 1.50% -0.23% 40 GNRC -1.31% 0.96% -0.91% -0.53% 2.57% -0.07% 0.37% -0.42% -1.32% 1.20% -1.80% 2.28% 2.38% 2.87% 2.77% 19.74% -0.53% 1.09% 12.01% -3.52% 0.30% PGTI -1.26% 0.50% 0.62% 2.71% -1.67% 0.67% -1.56% -0.63% 5.53% 0.99% 0.23% 0.62% -1.55% 0.32% -5.40% 1.17% -6.90% -6.19% 4.59% 3.53% -3.97% OC -0.34% -4.10% -0.68% 0.91% 1.20% 0.15% 0.60% 0.49% -0.29% 0.74% -0.10% 1.28% 0.72% -1.06% 2.51% -0.82% 2.09% -0.02% -0.99% -1.47% 1.51% XYL 0.28% -0.16% -0.30% -1.07% 1.57% -0.31% 1.57% -1.63% -1.41% 0.07% 0.30% 0.99% -2.83% 0.89% -0.41% 2.72% 3.48% 2.01% 1.17% -0.62% -0.41% Exhibit 4: Cumulative Average Abnormal Return (CAAR) (from t=-2 to t=2) t -2 -1 0 1 2 HD 0.14% -0.82% 0.43% -1.46% -1.46% LOW 0.01% -0.80% 0.18% -1.60% -1.43% BECN 0.76% 1.01% 4.04% 3.26% 4.76% SKY -0.61% 0.90% 1.00% 0.53% 0.53% WGO -0.04% 0.27% 2.27% 0.90% 2.41% BGG 0.38% 2.55% 4.68% 3.39% 5.40% FLS -0.12% 0.28% -0.23% 0.57% -0.04% t -2 -1 0 1 2 PNR 0.00% 1.13% 1.88% 0.35% 3.28% PWR 0.38% 1.38% 4.16% 5.25% 5.31% EME 2.15% 2.78% 2.86% 3.39% 2.03% MTRX -1.17% -1.14% -0.48% -1.81% 1.91% CLH 0.07% 0.36% 4.37% 4.10% 4.04% GNRC 2.87% 5.64% 25.38% 24.85% 25.94% PGTI 0.32% -5.09% -3.91% -10.82% -17.01% 41 GVA -0.26% 0.29% 2.06% 3.60% 5.77% OC -1.06% 1.45% 0.63% 2.73% 2.71% XYL 0.89% 0.48% 3.20% 6.68% 8.69% Exhibit 5: Graphical Representation of Rolling Cumulative Average Abnormal Returns Cumulative Average Abnormal Return (CAAR) 5.00% 4.00% 3.00% 2.00% 1.00% 0.00% -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 -1.00% -2.00% t CAAR (13 Securities Traded for All Hurricanes) CAAR (All 17 Securities) Exhibit 6: Graphical Representation of Average Abnormal Returns Average Abnormal Return (AAR) 3.00% 2.50% 2.00% 1.50% 1.00% 0.50% 0.00% -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 -0.50% -1.00% t AAR (13 Securities Traded for All Hurricanes) 42 AAR (All 17 Securities) 4 5 Exhibit 7: Graphical Representation of Rolling Cumulative Average Abnormal Returns Cumulative Average Abnormal Return (CAAR) (from t=-2 to t=2) 3.50% 3.00% 2.50% 2.00% 1.50% 1.00% 0.50% 0.00% -2 -1 0 1 2 t CAAR (13 Securities Traded for All Hurricanes) 43 CAAR (All 17 Securities) REFERENCES 1992: A Year of Catastrophes. (1993, February). Aide Magazine (USAA), p. 1. Retrieved March 7, 2017 Bennett, D. (2008, July 8). Do natural disasters stimulate economic growth? Retrieved March 8, 2017, from The New York Times: http://www.nytimes.com/2008/07/08/business/worldbusiness/08ihtdisasters.4.14335899.html Blau, Benjamin M.; Van Ness, Robert A.; Wade, Chip. (2008, December). Journal of Risk & Insurance, Vol. 75, Issue 4, 967-996. Retrieved from https://papers.ssrn.com/sol3/papers.cfm?abstract_id=960999 Brown, S. J., & Warner, J. B. (1980, September). Measuring security price performance. Journal of Financial Economics, 8(3), 205-258. Retrieved from http://dx.doi.org/10.1016/0304-405X(80)90002-1 Calkins, K. G. (2005, July28). Applied Statistics - Lesson 11 - Power and Sample Size. Retrieved March 13, 2017, from Andrews University: https://www.andrews.edu/~calkins/math/edrm611/edrm11.htm Cherwa, J. (2016, October 7). After Florida 'dodged a bullet' from Hurricane Matthew, South Carolina braces for major storm surge. Retrieved from Los Angeles Times: http://www.latimes.com/nation/la-na-hurricane-matthew-20161007-snapstory.html 44 CNN. (2016, August 23). Hurricane Katrina Statistics Fast Facts. Retrieved March 8, 2017, from CNN: http://www.cnn.com/2013/08/23/us/hurricane-katrina-statisticsfast-facts/index.html Dolce, C. (2015, October 18). 2015 Sets a New Record for Category 4 and 5 Hurricanes and Typhoons. Retrieved March 8, 2017, from The Weather Channel: http://www.nhc.noaa.gov/surge/ Event Study Metrics. (n.d.). Event Study Methodology. Retrieved March 8, 2017, from Event Study Metrics: http://eventstudymetrics.com/index.php/event-studymethodology/ Fama, E. F. (1970, May). Efficient Capital Markets: A Review of Theory and Empirical Work. The Journal of Finance, Vol. 25, No. 2, 383-417. Retrieved from http://www.jstor.org/stable/2325486 Fama, E. F., Fisher, L., Jensen, M. C., & Roll, R. (1969, February). The Adjustment of Stock Prices to New Information. International Economic Review, 10, 1-27. Retrieved from https://ssrn.com/abstract=321524 FEMA. (n.d.). Orleans Parish - Hurricane Katrina. Retrieved March 7, 2017, from FEMA: https://www.fema.gov/media-library-data/14399290922862741ee543b024fdab8b3bef80b835f3c/FS_Orleans_Final_081215.pdf Ferrell, J. (2008, September 19). Hurricane Ike Caused 7.5 Million Power Outages. Retrieved March 7, 2017, from AccuWeather: http://www.accuweather.com/en/weather-blogs/weathermatrix/hurricane-ikecaused-75-million-power-outages/16952 45 Frost, J. (2014, January 23). Regression Analysis: How to Interpret S, the Standard Error of the Regression. Retrieved March 7, 2017, from The Minitab Blog: http://blog.minitab.com/blog/adventures-in-statistics-2/regression-analysis-howto-interpret-s-the-standard-error-of-the-regression Fuqua School of Business, Duke University. (2017, January 10). Statistical forecasting: notes on regression and time series analysis. Retrieved March 8, 2017, from Fuqua School of Business, Duke University: https://people.duke.edu/~rnau/411home.htm Generac Investor Relations. (2013, February 14). Generac Reports Record Fourth Quarter and Full-Year 2012 Results. Retrieved March 7, 2017, from Generac: http://investors.generac.com/phoenix.zhtml?c=232690&p=irolnewsArticle&ID=1785023 Huffington Post. (2012, October 30). Sandy Is The Largest Hurricane To Ever Form In The The Atlantic Basin. Retrieved March7 2017, from Huffington Post: http://www.huffingtonpost.com/2012/10/30/hurricane-sandy-largesthurricane_n_2045163.html Kim, S. (2012, October 30). Stock Markets to Reopen After Hurricane Sandy Shut Down NYSE. Retrieved March 7, 2017, from ABC News: http://abcnews.go.com/blogs/business/2012/10/stock-markets-to-reopen-afterhurricane-sandy-shut-down-nyse/ Koerniadi, H., Krishnamurti, C., & Tourani-Rad, A. (2012). Natural Disasters – Blessings in Disguise? Financial Markets & Corporate Governance Conference. Retrieved from https://ssrn.com/abstract=1969593 46 LaCommare, K. H., & Eto, J. H. (2004). Understanding the Cost of Power Interruptions to U.S. Electricity Consumers. University of California Berkeley, Energy Analysis Department. Ernest Orlando Lwrence Berkeley National Laboratory. Retrieved March 11, 2017, from https://emp.lbl.gov/sites/all/files/lbnl-55718.pdf Lamb, R. P. (1995, March). An Exposure-Based Analysis of Property-Liability Insurer Stock Values Around Hurricane Andrew . The Journal of Risk and Insurance, 62(1), 111-123. Retrieved from http://www.jstor.org/stable/253695 Liu, H., Ferreira, S., & Karali, B. (2015). Hurricanes as News? A Comparison of the Impact of Hurricanes on. Southern Agricultural Economics Association Annual Meeting. Atlanta, Georgia. Retrieved March 8, 2017, from http://ageconsearch.umn.edu/bitstream/196845/2/Hurricanes%20as%20News_FI NAL_1152015.pdf NOAA. (2016, March). What is a hurricane? Retrieved March 8, 2016, from National Ocean Service: http://oceanservice.noaa.gov/facts/hurricane.html NOAA. (n.d.). Saffir-Simpson Hurricane Wind Scale. Retrieved March 8, 2017, from National Hurricane Center: http://www.nhc.noaa.gov/aboutsshws.php NOAA. (n.d.). Storm Surge Overview. Retrieved March 8, 2017, from National Hurricane Center: http://www.nhc.noaa.gov/surge/ Nordhaus, D. W. (2006, December). The Econonics of Hurricanes in the United States. Annual Meetings of the American Economic Association. The National Bureau of Economic Research. Retrieved from http://www.nber.org/papers/w12813 Ocean Tides. (n.d.). Barometric Perssure and Hurricanes. Retrieved March 11, 2017, from Sciencing: http://sciencing.com/barometric-pressure-hurricanes-22734.html 47 Ofek, E., & Richardson, M. (2000, January). The IPO Lock-Up Period: Implications for Market Efficiency And Downward Sloping Demand Curves. rn School of Business, New York University. Retrieved March 13, 2017, from http://people.stern.nyu.edu/eofek/lockup.pdf Pasch, Richard, J., Blake, S. E., Cobb III, H. D., & Roberts, D. P. (2006, January 12). Tropical Cyclone Report, Hurricane Wilma. NOAA. Retrieved March 7, 2016, from NOAA. Perold, A. F. (2004). The Capital Asset Pricing Model. The Journal of Economic Perspectives, 3-24. Retrieved March 9, 2017, from http://www.jstor.org/stable/3216804 Rice, D. (2016, October 10). Hurricane Matthew dumped enough rain to fill the Rose Bowl 163,000 times. Retrieved March 8, 2017, from USA Today: http://www.usatoday.com/story/weather/2016/10/10/hurricane-matthew-by-thenumbers/91856938/ Robinson, C. J., & Banwayo-Skeete, P. (2016, March). The Financial Impact of Natural Disasters: Assessing the Effect of Hurricanes & Tropical Storms. 1-21. Retrieved March 8, 2017, from https://ssrn.com/abstract=2845429 The University of Queensland. (n.d.). Event Study Methodology. Retrieved March 7, 2017, from http://teach.business.uq.edu.au/courses/FINM6905/files/module3/notes/event%20study%20methodolgy.pdf Tribune Media Wire. (2016, March 18). The 15 most destructive hurrcanes since 1950. Retrieved March 8, 2017, from Fox8: http://myfox8.com/2016/03/18/the-15most-destructive-hurricanes-since-1950/ 48 U.S. Department of Commerce. (2013). Economic Impact of Hurricane Sandy: Potential Economic Activity Lost and Gained in New Jersey and New York. U.S. Department of Commerce. Retrieved March 7, 2017, from http://www.esa.doc.gov/sites/default/files/sandyfinal101713.pdf U.S. Global Change Research Program. (n.d.). Changes in Hurricanes. Retrieved March 8, 2016, from GlobalChange.gov: http://nca2014.globalchange.gov/report/ourchanging-climate/changes-hurricanes University of St. Thomas. (n.d.). WRDS: Wharton Research Data Services: CRSP Basics. Retrieved March 7, 2017, from University of St. Thomas: http://libguides.stthomas.edu/c.php?g=88674&p=572679 Winter, C. (2012, November 5). Hurricane Sandy Boosts the Generator Makers. Retrieved March 8, 2017, from Bloomberg: https://www.bloomberg.com/news/articles/2012-11-05/hurricane-sandy-booststhe-generator-makers 49 50