The Determinants of Residential Property Damage Caused by Hurricane Andrew
Journal article by Paul Fronstin, Alphonse G. Holtmann; Southern Economic Journal, Vol. 61, 1994
Journal Article Excerpt
The determinants of residential property damage caused by Hurricane Andrew. by Paul Fronstin , Alphonse G. Holtmann I. Introduction On the morning of August 24, 1992, Hurricane Andrew demolished the southeastern part of Florida. Hurricane Andrew was the costliest natural disaster in U.S. history: fifteen deaths were directly attributable to the storm, 175,000 individuals were left homeless, 25,000 homes were completely destroyed, and another 100,000 were damaged. Over 7,800 business were affected, comprising 14 percent of Dade County's economy. Costs have been estimated at $20 billion in damages, $10 billion in cleanup, $15.5 billion in insurance claims, and $1.04 billion in agriculture loss. In comparison, preliminary estimates of costs from the recent California earthquake have reached $20 billion in damages and $5 billion in insurance claims. While both natural disasters cause widespread damage, the nature of the damage is different and affects private and public decision making differently.(1) The strength of a hurricane can be measured by its wind speed. In the strongest part of the storm sustained wind speeds were estimated to be above 133 miles per hour (mph).(2) Wind gusts reached over 175 mph in other areas, yet some of the most damaged areas were not in the strongest part of the storm. For example, a well known subdivision of homes named Country Walk only experienced sustained winds of 97 mph or less, with gusts reaching 127 mph, but all the homes were completely destroyed. Surprisingly, other subdivisions close to Country Walk were not as severely damaged: a subdivision just eight blocks south of Country Walk, and closer to the eye of the hurricane, had very few homes destroyed or severely damaged. A common pattern of destruction seems to emerge from analysis of the storm: older homes incurred less damage than newer homes. Investigations by expert hurricane scientists and engineers came to the conclusion that wind speed was not the only factor that caused severe destruction. They believe that low quality construction, faulty designs, and flimsy materials--common problems in newer dwellings--all played a role in the severity of the damage. These problems can partly be attributed to an eroding building code. Although Dade County Commissioners adopted the South Florida Building Code in 1957, a code considered to be the strongest code in the nation, Table I shows that during the past 35 years the Dade County Board of Rules and Appeals allowed building standards to become less stringent. For example, in 1961, just four years after adopting the building code, the Dade County Board of Rules and Appeals approved the use of power driven staples for roof construction. Houses could be built more quickly because roofers did not have to use hand driven nails. Additional changes in the building code occurred during the 1970s and 1980s. In 1984, the building code was changed in order to allow the use of waferboard (pressed wood) on roofs instead of plywood. In this paper, we offer two explanations for the nature of the hurricane damage. First, our empirical analysis shows that the erosion of the building code resulted in more damage to newer homes than to older homes. Second, we hypothesize that consumers have substituted home-owners insurance for structurally sound homes that are built to withstand hurricanes because of the rising cost of building a home relative to the cost of home-owners insurance. This is the common problem of moral hazard. Data limitations confine our ability to test this hypothesis. Our empirical specification utilizes a two-limit Tobit model to explain the effects of wind speed, quality of housing, and recent changes in building codes in southeast Florida on the probability of incurring substantial housing damage from Hurricane Andrew. II. The Model In the wake of Hurricane Andrew, the community of southeast Florida, and others perhaps, began to ask whether the appalling physical damage was avoidable. Did builders use inferior materials for newly constructed houses and condominiums? Did government officials and inspectors neglect their duty in requiring and enforcing building codes? Did consumers support changes in the construction industry? Finally, are we all becoming more slipshod in what we do? There appears to be at least some evidence that the answer to all of the above questions is yes. But, if the answer is yes, why is this so? There is some evidence that older, presumably better built houses, received less damage from Hurricane Andrew than did newer houses, adding support to the notion that housing quality is not what it used to be--at least in terms of combating a hurricane. It is possible that increased private and public insurance of property, and greater ease of protecting one's life from hurricanes have increased the moral hazard associated with protecting our housing property causing increased demand for lower quality housing. This is supported by the fact that government officials allowed the South Florida Building Code to be weakened during the past 35 years, allowing builders to substitute less expensive inputs. This is also supported by the fact that modern improvements in predicting hurricane behavior, though not perfect, have improved dramatically over time, and methods of communication and transportation have also improved, reducing the cost of evacuation. The alternative to evacuation is building a more costly hurricane resistant structure. This idea of the changing costs of protecting one's self from a hurricane is consistent with the fact that property damage from Hurricane Andrew was very high, but loss of life was remarkably low. Consumers may have contributed to the damage as well. The building code may have been relaxed in response to consumer demands. Cheaper inputs mean lower building costs and lower priced homes. It is possible that consumer preferences have changed in such a way that consumers are willing to substitute product characteristics at the expense of solid construction. It is also possible that technological innovation in building techniques since World War II contributed to the damage. Innovation that would make houses cheaper to build may not be built with hurricanes in mind? For example, wood framed houses suffered more damage than concrete-block houses. Along with the changing relative costs of assuring protection of one's life, limb, and human capital, there has been a growing market for home-owner's insurance, stemming from the recent requirement by lending institutions that mortgaged properties be insured. This insurance, like many types of insurance, reduces the homeowner's and the mortgage holder's incentive to invest in costly procedures to ensure the house's resistance to threats from hurricane damage. Insurance companies have a right to require that a person's house meet certain standards before insuring the home, but there is some evidence that insurance companies did not inspect homes prior to insuring them.(4) Lastly, over time various forms of social insurance have become more available. Federally guaranteed low interest loans for rebuilding property damaged by a hurricane and national guard protection of unprotected property confer a number of important benefits on people in the storm area, but this social insurance also reduces the incentive for individuals to remain in a threatened area to maintain and protect their property during a hurricane.(5) In formulating the model to estimate the determinants of property damage due to a hurricane, the following model is estimated: [y.sub.i] = f(H, S) (1) where y measures the percentage of the homes that were destroyed in subdivision i, H is a vector of the characteristics of the home, and S is a vector of the characteristics of the hurricane. The variables included in vector H consist of characteristics reflecting the homes ability to withstand the hurricane's strength. The price of the home is included to represent the cost and quality of the home, and the age of the home is included to control for government regulations, in effect at the time the home was built, that should also affect the quality of the structure. Assuming that price is a proxy for quality, we expect a negative relationship between the damage to a home and the price of the home. Our a priori expectations of the relationship between the homes age and the homes damage is also negative. This expectation is based on the rules of the South Florida Building Code, which have eroded, as shown in Table I. The variables included in the vector S consist of hurricane characteristics which are included to control for the power of the hurricane. We measure the characteristics of the hurricane by including variables to control for the wind speed of the storm and whether or not a home was subject to severe coastal flooding. III. Data and Econometric Specification The data used in this study were tabulated by the National Oceanographic and Atmospheric Administration (NOAA) and published in a special report by The Miami Herald, on December 20, 1992. The report contains data on 420 subdivisions or condominium developments in southeast Florida. Subdivisions were only included if they contained at least 25 homes, and at least 25 percent of the homes had been inspected as of mid-November 1992.(6) The data provided by the Miami Herald include the number of homes in the development, the percent of the homes that were judged to be uninhabitable after the storm, the average year of construction of the homes in the development, the average assessed value of the homes, the estimated sustained wind speed that the development was subjected to during the storm, and whether or not the development was subjected to the storm surge.(7) Even though we are using average data from subdivisions and condominium developments, each observation represents the average of a fairly homogeneous group of homes. For example, all of the homes within a given subdivision were often built by the same developer at approximately the same time.(8) In order to predict the determinants of property damage due to Hurricane Andrew, we estimate a two-limit Tobit model [5], corrected for the fact that the dependent variable, percentage of houses uninhabitable within a development, had different weight depending on the size of the housing development and the number of homes inspected after the storm. The Tobit model is an appropriate estimating technique because it allows us to constrain our estimated predictions to zero percent as a possible lower limit and one hundred percent as a possible upper limit. The Tobit model is given by the following equation: E(y [where] x, L [is less than] y* [is less than] U) = [Beta]x + [Sigma] ([[Phi].sub.L] - [[Phi].sub.U])/([[Phi].sub.U] - [[Phi].sub.L]) (2) where [[Phi].sub.j] = [Phi][(j - [Beta]x)/[Sigma]], j = L, U, [[Phi].sub.j] = [Phi][(j - [Beta]x)/[Sigma]], j = L, U, and y equals y* if y* falls in some range, otherwise y equals the limit values. [Phi] represents the density function of the standard normal, [Phi] represents the cumulative distribution function of the standard normal distribution, [Sigma] represents the standard error of the estimate, and L and U represent the lower and upper limits of the distribution, respectively. Predictions from the model are obtained from the unconditional mean: E[y] = L[[Phi].sub.L] + U (1 - [[Phi].sub.U]) + ([[Phi].sub.U] - [[Phi].sub.L]) [Beta]x + [Sigma]([[Phi].sub.L] - [[Phi].sub.U]). (3) In addition, the following equation is used to estimate the marginal effects of the model: [Delta]E[y]/[Delta][x.sub.i] = [[Beta].sub.i] ([[Phi].sub.U] - [[Phi].sub.L]). (4) In our model, the dependent variable is the percentage of homes in a particular subdivision that were declared "uninhabitable". Approximately ten percent of the subdivisions were completely destroyed. Only fourteen percent of the subdivisions incurred no severe damage. The explanatory variables in the model include the average assessed value of the property (VALUE). The value of the home can be used as a proxy for the quality of the home. We expect higher priced homes to be better built than less expensive homes; therefore, we predict a negative relationship between the amount of damage and the price of the home. Dummy variables to control for three categories of sustained wind speed are also included in the model. These categories include a group that represents sustained wind speed less than 97 mph (the base group), a category for wind speeds between 97 and 114 mph (WS2), and a category for sustained wind speeds above 115 mph (WS3).(9) We expect higher wind speeds to have a greater impact on the amount of damage in a development. There has been some controversy pertaining to the actual sustained wind speeds and wind gusts during the hurricane. Some estimates of wind speed are actually much higher than the ones we use in this study. Government researchers have estimated wind speeds much stronger than previously believed [3]. Alternatively, Professor Ted Fujita, of the University of Chicago, asserts that factors such as microburts and mini-swirls of wind contributed more to the damage than the actual sustained wind speed [1]. Nevertheless, scientific conclusions are being heavily disputed in scientific circles. For the purposes of our study, we measure the strength of the storm using dummy variables. Since we are only concerned with relative wind speed, as opposed to actual wind speeds, any potential measurement error in wind speed should not affect our results.(10) The age of the home is also controlled for in this analysis. Age (AGE) and age-squared (AGESQ) are included to control for a possible non-linear relationship between the age of the development and the damage to the homes in the development. As mentioned previously, because of changes in the building code, we expect that older homes incurred less damage than newer homes. Our model also includes interaction terms between the sustained wind speed variables and the age variables. These interaction terms are included to separate out the effects of wind speed on homes of different ages. Finally, a dummy variable is also included to control for those developments that were located within the storm surge area (SURGE). These homes are subjected to the possibility of flooding, thus we expect these homes to incur a greater amount of damage than homes located a greater distance inland. IV. Empirical Results The results from our empirical analysis are presented in Table II and conform to our prior expectations. There is strong support for the hypothesis that the assessed value of a home (VALUE) is an indication of the quality of the construction of the home. Our analysis shows that an increase in the average assessed value of a home within a subdivision has a significant negative effect on the amount of damage incurred in the subdivision, but the magnitude of the marginal effect is small. For instance, using the marginal effect of -0.00011 from equation (4) of the Tobit model,(11) a $10,000 increase in the average assessed value of a home within a given subdivision will result in a decrease of 1.1 percent of the homes that were uninhabitable after the storm. Even though the effect of the average assessed value of a home on the damage incurred is relatively small in magnitude, our results do suggest that the cost of a home is a significant indicator of the quality of the home with respect to the home's ability to withstand the force of a hurricane. Table II. Estimated Coefficients and Means(a)Figure 1 contains the damage predictions from the model based on wind speed and the year that the development was built. The trends exhibited in Figure 1 lead us to believe that the erosion of the building codes in southeast Florida was at least partially to blame for the poor performance of newer subdivisions. This trend is especially pronounced for subdivisions located within the lowest wind speed category. We can see from the graph that subdivisions built in the late 1960s incurred the least amount of damage, but after 1970 the percentage of uninhabitable homes significantly increased within the lowest wind speed category and begins to converge with the highest wind speed category. This suggests that because the building code became less stringent very new subdivisions suffered a large amount of damage regardless of the wind speed during the hurricane. The dummy variables in Table II representing different wind speeds indicate that very new subdivisions located within the lowest wind speed suffered almost as much damage as very new subdivisions located within the highest wind speed category. This trend is also supported by the predictions in Figure 1, which shows that the percentage of homes completely destroyed within a subdivision converges to nearly one-hundred percent for newer homes within both the lowest wind speed category and the highest wind speed category. There is little evidence that the storm surge had an effect on the amount of hurricane damage. We are not surprised that the storm surge variable is not significant because only three homes were completely destroyed by the surge.(12) Table III contains the predicted percentage of uninhabitable homes within a subdivision based on the subdivisions age and the sustained wind speed that the homes were subjected to during the storm. These results illustrate the true picture of destruction due to Hurricane Andrew. For example, a subdivision built 5 years prior to Hurricane Andrew had 72.21 percent of the homes in it completely destroyed if it was only subjected to wind speeds less than 97 mph, but in a subdivision that was subjected to wind speeds greater than 115 mph, 96.65 percent of the homes were completely destroyed. The results are consistent with the fact that relatively new homes in areas of higher sustained winds have a higher probability of incurring severe damage than older homes. As we mentioned earlier, the most striking result is that newer homes were not able to resist winds below 97 mph. Nevertheless, the highest wind speed increases the risk of damage regardless of the age of the subdivision, as we would expect. Table III. Predicted Percentage of Uninhabitable Homes within a Subdivision,Our results are reinforced by the contents of Figure 2. In Figure 2, the probability function of a given home being totally destroyed is presented, based on the homes age and the wind speed that the home experienced during the hurricane. Again, the general trends in Figure 2 suggest that newer homes have a higher probability of being totally destroyed than older homes. The figure also suggests that homes located within the highest wind speed category have the highest probability of being destroyed, but the gap between the highest wind speed and the lowest wind speed closes as the homes become newer. Using the predictions of the model, we can calculate the costs attributable to low quality construction by simulating the model with the assumption that all of the developments were built in the late 1960s, the period of time with the least amount of damage.(13) The homes in our sample represent approximately $1.1 billion in damages. We estimate that $376 million (33 percent) would have been saved if all of the developments were built like the homes in the late 1960s. This is a lower bound estimate. It only takes into account the costs to homes that were completely uninhabitable. Homes that were severely damaged, but inhabitable after the hurricane were not included in the calculation.(14) If we knew the costs of inspecting the homes and the benefits that accrued to homeowners, because of the relatively lower priced homes, we could calculate the marginal value of lower quality construction and the erosion of the building code that would have accrued to consumers since the late 1960s. V. Conclusions and Policy Implications Our analysis of published data support the emerging conclusion that newer houses (those built after the 1960s) sustained a greater amount of damage from Hurricane Andrew than did houses built earlier in the period, ceteris paribus. We find that houses that were subjected to higher sustained wind speeds suffered more damage, though the pattern is ambiguous for the lower two wind speed categories. Lastly, the assessed value of the house is inversely related to the probability of the home being uninhabitable after the storm, suggesting that higher prices are an indication of higher quality. We give a number of reasons for the large amount of destruction in the newer homes. Building codes appear to have eroded over time, which may be attributed to a failure of government to represent homeowners interests, or to a change in the value that homeowners placed on combatting hurricane damage because of changes in various types of insurance. Asymmetric information between buyers and sellers is a problem in the home construction industry, but building codes are meant to correct this failure in the market. Alternatively, the building code may have been relaxed in response to consumer demands. Cheaper inputs mean lower costs of building a house. It is also possible that consumer preferences have changed in such a way that they are willing to substitute product characteristics at the expense of solid construction. If homes were built using the standards of the late 1960s, there would have been at least a 33 percent savings in the amount of damage.(15) In addition, we also suggest that the deterioration of standards for protecting property from a hurricane may revolve around nothing more than the problem of moral hazard associated with many types of insurance. Consumers who knew their property was insured and who were able to evacuate some of the areas affected by the storm were more able and willing to leave their property than would have otherwise been the case. Although the growth in insurance of various types allowed people to reduce the protection of their property, it increased the incentive to protect their lives. Unfortunately, due to the lack of data, we can not test this hypothesis. One puzzle that remains is why insurance companies underwrote policies with such complete coverage in the face of growing problems of deteriorating housing code protection. According to industry experts, insurance companies simply made a mistake due to their lack of experience with hurricanes of great magnitude. Since the hurricane, thirteen insurance underwriters have completely withdrawn from the market for homeowners insurance and others have simply put a cap on the number of new policies that can be written. As a result, the Florida Department of Insurance has created a wind storm pool in Dade and Broward Counties. This wind storm pool would insure people against the risk of hurricanes when they cannot get insurance from a private company. Insurance companies would then be more inclined to offer basic homeowners insurance if hurricane risks were removed by the wind storm pool. 1. Hurricanes tend to cause more damage to private residences and businesses, while earthquakes tend to cause more damage to public infrastructure, such as bridges and highways. In the hurricane, 125,000 homes were destroyed or damaged, compared with 25,000 damaged or destroyed homes due to the earthquake. 2. At the National Hurricane Center in South Florida, wind speed reached 164 mph before the measuring instruments were blown off of the building. 3. Some homes fared worse than others because of the design of the building. For example, homes with "gable end" style roofs suffered more damage than homes with other types of roofs. 4. A front page article in the New York Times on December 28, 1993 [4], quotes Edward Young, senior vice-president at Allstate Insurance, as saying that insurance companies assumed that government officials were enforcing building codes. The article also states that insurance companies are going to begin inspecting the homes themselves to determine premiums, instead of relying on general risk guidelines based on location and construction type. 5. Individuals can protect their property during a hurricane by boarding up windows or holding mattresses against windows after they have broken. More importantly, individuals can protect their property from additional losses after a hurricane by remaining on their property to guard against vandalism. It is more difficult to return to a hurricane stricken area then it is to stay in the area because of road congestion caused by road obstructions. 6. Trailer parks were not included in this study. They represent an inherently different commodity in the housing market and virtually all of the trailers in the strongest part of the storm were destroyed. 7. The storm surge is the flooding along coastal areas created by the hurricane. The surge usually arrives an hour or two before the strongest part of the hurricane reaches land. The stronger the storm, the further inland the surge reaches. The area of the greatest storm surge is in the upper eye wall and reached approximately 16 feet during Hurricane Andrew. Surprisingly, only three homes were completely destroyed by the storm surge. 8. Though we do not have statistical evidence concerning the homogeneity of houses, close observations of developments reveals that most developments of houses contained many houses of the same exact style. 9. We are able to distinguish between two other categories of wind speed--between 115 and 133 mph, and above 133 mph--but initial estimates of our model suggested that there was no significant difference between these two categories. We attribute this to the fact that damage was severe for all homes which experienced sustained wind speeds above 115 mph. Essentially, higher wind speeds do not cause more damage than lower wind speeds once the lower wind speed reaches a certain level, in this case 115 mph. 10. We estimated our model using wind speed measures given in Powell, Houston, and Reinhold [3]. The basic results do not change and are available from the authors upon request. 11. The marginal effect (-0.00011) is calculated using the following equation: [[Beta].sub.i]([[Phi].sub.U] - [[Phi].sub.L]) = [[Beta].sub.i] * ([Phi] ((100 - [Beta]x)/[Sigma]) - [Phi] ((- [Beta]x)/[Sigma])), where [[Beta].sub.i] (= -0.00013) is the Tobit coefficient, [Beta] represents the vector of estimated coefficients, [Phi] represents the cumulative distribution function of the standard normal distribution, [Sigma] (=35.55) represents the standard error of the estimate, and x is the vector of explanatory variables evaluated at the mean. 12. We are, however, surprised that only three homes were destroyed by the surge, which reached over ten feet in many areas. Additionally, many homes suffered quite a bit of damage from the storm surge, but none of these homes were declared "uninhabitable" after the storm. 13. We chose 1968 for the calculation. 14. We only know the percent of the homes in a development that were completely uninhabitable after the hurricane. We do not know the extent of the damage to homes that were inhabitable after the hurricane. Developments that were not included in the sample were also not included in our calculation. 15. This savings would be offset by the additional costs of inspecting the homes, and the benefits that accrued to consumers from having lower priced homes in the interim. References 1. Dorscher, John, "Grasping Andrew's Full Fury." Miami Herald, 22 July 1993. 2. Getter, Lisa, "Building Code Eroded Over Years." Miami Herald, 11 October 1992. 3. Powell, Mark D., Sam H. Houston, and Timothy A. Reinhold. "Standardizing Wind Measurements for Documentation of Surface Wind Fields in Hurricane Andrew." Paper presented at the ASCE Conference on Hurricanes of 1992, Miami, Florida, 1-3 December 1993. 4. Quint, Michael, "Insurers Maneuver to Cut Their Risks from Large Storms." New York Times, 28 December 1993. 5. Rosett, Richard N. and Forrest D. Nelson, "Estimation of the Two-Limit Probit Regression Model." Econometrica, January 1975, 141-46. Table I. History of the South Florida Building Code 1961 Dade County Board of Rules and Appeals gave approval for power driven staples for roof installations. 1970 County allowed Masonite hardwood siding to be allowed on exterior walls of houses, without plywood to back it up. 1973 Professional engineers of Dade County tell board they can't guarantee a building is properly constructed 1979 Special inspectors refused to sign affidavits that a building was constructed according to code unless the forms were qualified by the words "to the best of my knowledge and belief." The board acquiesced and weakened the code with the additional clause. 1980 Board approved the use of thinner roofing felt because it would save consumers money, due to increase in oil prices. 1982 Board allowed the building of cost-effective two story wood framed condominiums that violated the code's standards regarding non combustible load bearing exterior walls. The code change allowed the builder to construct wood partitions instead of concrete block walls. 1983 Board learns that few fasteners on market would pass full blown wind load test. Hears that staples fare worse than nails. 1983 Board learned that asphalt shingles can only resist top speed of 63 mph. Despite acknowledgements from manufacturers about the limitations of the shingles, board refused to outlaw their use. 1984 Board allowed construction of entry doors to open outward even though it was technically in violation of the code. Board wanted to maintain the "standard of the industry and the design custom of the community." 1984 Board allowed waferboard to be used instead of plywood on roofs. 1984 Consulting engineer tells board that roofers say staples are not working. 1989 Consultant reporting on Hurricane Hugo damage concludes that workmanship in regards to roofs may need to be emphasized in Dade. Source: Miami Herald [2]. -1- |
End of free preview...
To continue reading this publication, you must have a Questia Subscription.
Try Us Today! Click Here
Questia provides the world's largest online library of scholarly books and journal articles, with integrated footnote and bibliography tools, highlighting, note taking and book marking. With a Questia subscription, you'll have access to the full text of more than 67,000 books and 1.5 million articles.
Already a subscriber? Login:
Sponsored Links
Free Newsletter
Get helpful how-to's, writing tips, search strategies, quizzes & more!
Search the Library
