Review on Energy Resilience

A Time Series Approach to Examining Regional Economic Resiliency to Hurricanes

By BRADLEY T. EWING*, DAAN LIANG†, and YUEPENG CUI‡

ABSTRACT. Hurricanes disrupt business processes and activities, energy distribution and consumption, and services of infrastructure and lead to a reallocation of resources and their uses. This research models the relationship among economic and engineering measures of the state of the built environment in order to provide insight into a region’s ability to withstand and recover from a future hurricane. As such, we provide a quantitative approach to modeling and meas- uring a region’s economic resiliency. The findings have implications for developing strategies for long-term sustainability of economic regions.

Introduction

Hurricanes disrupt business processes and activities, energy distri- bution and consumption, and services of infrastructure, and lead to a reallocation of resources and their uses temporarily and possibly permanently (Chow and Elkind 2005; Kaisera et al. 2009; Comfort and Haase 2006). Thus, models that incorporate traditional economic measures with various engineering measures of the state of the built environment may provide insight into a region’s ability to withstand and recover from a future hurricane. In particular, our premise is that

*Corresponding author. Rawls Professor of Energy Economics, Rawls College of

Business and National Wind Institute, Texas Tech University; e-mail: bradley.ewing@

ttu.edu; Phone: 806-843-3939.

†Associate Professor, National Wind Institute and Department of Construction Engi-

neering & Engineering Technology, Texas Tech University.

‡Doctoral Student in Wind Science and Engineering, National Wind Institute, Texas

Tech University.

Acknowledgments: This material is based upon work in part supported by the

National Science Foundation under Grant No. CMMI 1131392 & 1000251. Any opinions,

findings, and conclusions or recommendations expressed in this article are those of the

authors and do not necessarily reflect the views of the National Science Foundation.

American Journal of Economics and Sociology, Vol. 73, No. 2 (April, 2014). DOI: 10.1111/ajes.12071 © 2014 American Journal of Economics and Sociology, Inc.

an economy that has a stable relationship among these performance factors would likely be more resilient and sustainable over time.

The goal of this research is to develop a quantitative time series model for assessing regional economic resilience to hurricane hazards. The results will support decision making in disaster recovery and mitigation and should help produce long-term systemic benefits to communities at large. This article focuses on the greater Houston area, which has seen and continues to experience considerable hurricane- related impacts. In particular, we search for a common factor among broad measures of regional economic and engineering performance, namely, employment, energy, and the built environment. A common factor may be thought of as an underlying driving force that the key elements of the system tend towards over time, thus making the system stable and predictable.

In statistical terms, the existence of a common factor indicates that these measures are cointegrated or linked together in the long term. According to Engle and Granger’s (1987) representation theorem, a cointegrated system of variables ordered in the time domain may be modeled using a vector autoregression error correction model (VECM). From the VECM it is possible to discover the long-run error pattern and to formulate the adjustment process from disequilibrium. It is natural to think of a hurricane event as bumping (or shocking) a system from operating in equilibrium to a state of disequilibrium. If the system does not return to a long-run stable equilibrium, then the system is not cointegrated and therefore warrants governmental actions. However, for the case when cointegration is present, the system does return to equilibrium through some adjustment process and demonstrates the quality of self-resiliency. The VECM provides key information as to (1) which element(s) do the adjusting and (2) how long the adjustment process will take. Thus, this research pro- vides a quantitative approach to modeling and measuring a region’s economic resiliency. Moreover, the findings have implications for developing strategies for long-term sustainability of economic regions.

The article proceeds as follows. In the next section we describe economic resiliency in more detail and highlight some of the existing literature in this area of research. The third section discusses the Houston area and provides descriptions of Hurricanes Ike and Rita,

370 The American Journal of Economics and Sociology

both of which occurred during our study period and may have impacted or put at risk the greater Houston area in some way or another. The fourth section describes the data used in this study, followed by a description of the methodology employed, before presenting the empirical findings. The article concludes with some remarks about the usefulness of the model for evaluating the ability of communities to recover from disasters.

Overview of Economic Resiliency and Related Literature

While hurricane intensity and size are two important atmospheric factors for predicting damages and subsequent recovery (Zhang and Wang 2003; Irish et al. 2008), recent research in economics and engineering identifies a number of other elements that have significant influence on the health of economic enterprises (i.e., individual busi- nesses, industries, and regional economic activities) (Guimaraes et al. 1993; Ewing, Kruse, and Thompson 2005a; Taskin and Lodree 2010). In particular, the economic effects of hurricane-induced physical damage continue beyond landfall, a feature not accounted for in current weather-based recovery prediction models. The performance of an economic enterprise may or may not return to normal following a hurricane and even if it does, the process may require many months or years.

There are several reasons why we expect that blending engineering and economic-based data will enhance our understanding of resil- iency and recovery.

First, traditional business production models depend on resources such as human capital, working capital, technology, and physical assets (Hunt 1995; Barron et al. 2006). The availability of these resources may be altered by a disaster; however, the magnitude and duration of such changes are not yet well understood. While sig- nificant progress was made in the last decade in modeling struc- tural performance of building components and systems subject to extreme wind loading, notably through the development of HAZUS- MH,1 the knowledge gap between building damage and business interruption/performance is still being investigated (Greenberg et al. 2007). Moreover, it has been found that business interruption

Economic Resiliency to Hurricanes 371

following hurricanes is a common reason for business failure (Saleem et al. 2008).

Second, businesses are highly networked through supply chains inside and outside the boundary of hurricanes, which means that simply aggre- gating losses reported at the individual business level will likely under- estimate the total economic impact (Tang 2006; Lodree and Taskin 2009).

Third, the economic environments in which businesses operate are complex and transient, and determining how to separate hurri- cane impact from the normal business cycle is a great challenge (Dahlhamer and Tierney 1996; Guimaraes et al. 1993; Webb et al. 2000). As such, we expect that more explicit recognition of the interaction and dynamic behavior of engineering and economic-based variables will improve resiliency and recovery modeling, and shed light on what drives regional economic resiliency.

To date, econometric studies have identified several general classes of variables that explain differences in the rate and stability of recov- ery and economic performance in connection with natural disasters (Ewing et al. 2003, 2004, 2005a, 2005b, 2006, 2007, 2009; Ewing and Kruse 2002). These general classes include basic atmospheric charac- teristics of the windstorm (wind speed, central pressure, total rainfall, and duration), community vulnerability and infrastructure, state and federal assistance, and industrial/socioeconomic makeup (including supply chain) of the study area.

A RAND study suggested that a key factor determining how quickly New Orleans could be repopulated after Hurricane Katrina was the availability of housing (McCarthy et al. 2006). The faster housing becomes available, the faster people can return to cities, services, employment, and schools. Damage to housing stock was considered one of the most appropriate variables to estimate the reduction of population after a disaster (Plyer et al. 2010). Similarly, critical infra- structure systems, including water, wastewater, power and energy, transportation, and telecommunications, facilitate economic growth and provide the essential services of a modern society (National Research Council 2009). Therefore, accurate and continual assessment of the condition of housing stock and critical infrastructure systems will be of vital importance to both short-term relief effort and long- term recovery planning.

372 The American Journal of Economics and Sociology

A damage survey following a natural disaster is often conducted by engineers and researchers who are most interested in understanding how structures failed. For example, numerous engineering teams were deployed after Hurricane Katrina in 2005 to examine the performance of various structures including residential and commercial buildings, fire stations, airports, bridges, and levees (Liang et al. 2005; FEMA 2006; Womble et al. 2006; O’Connor and McAnany 2008; Mosqueda and Porter 2007; Marshall 2006; NIST 2006; ASCE 2007). These surveys, though extremely useful to the engineering and scientific community, were focused on a relatively small sample of failed structures and therefore may not be used to extrapolate the condition of the entire system.

Based on the review of the literature in engineering and econom- ics, we examine three broad categories of factors underlying the performance of an economy. First, local area employment provides a measure of labor market conditions and insight into output per- formance as firms first adjust employment relative to physical capital with changes in demand. Various labor market measures have been used in studies of the effects of natural disasters; however, the most common variable tends to be employment (Jeffrey and Anne 2008; Vigdor 2008; Ewing et al. 2005b, 2009). A measure of critical infra- structure systems and, specifically, energy use is represented by a proxy of retail gasoline price, which captures the influence of trans- portation costs on businesses and consumers. Transportation costs are known to fluctuate in reaction to natural disaster (Lewis 2009; FTC 2006; Deltas 2008; Boin and McConnell 2007). Additional engineering-related aspects of economic performance are measured by the use of new building permits, which represent construction, building innovation, infrastructure, and changes in the stock of the built environment (Stephen et al. 2007; Tatum and Terrell 2012; Tatano et al. 2004).

A key element in our research is to develop a model based on the Houston area that can be expanded to other hurricane-prone areas. In order to accomplish this goal, we focus on a feature known as parsimony. A model is said to be parsimonious when it has as few variables as possible but is still capable of capturing the system dynamics. Generally speaking, parsimony in econometric

Economic Resiliency to Hurricanes 373

modeling is desirable from a statistical perspective and makes eco- nomic models tractable (Enders 2009). Moreover, a parsimonious model allows us to extend the model structure to other regions and areas, as the data sources and series are comparable, readily avail- able, regionally specific, and released on a regular basis by reliable sources such as the Bureau of Labor Statistics, Energy Information Agency, etc.

We turn now to the descriptions of Hurricanes Rita and Ike, both of which occurred during the time period of our study and within a reasonable proximity to the greater Houston area and region. Speci- fically, we account for both of these hurricanes in our analysis as previous findings have shown that even close, nondirect hits can significantly impact on an economy. These impacts and associated risks may be attributed to a number of factors, including expectations of price spikes, demand surge, and supply disruptions (Ewing et al. 2005a). Moreover, natural disasters often have effects permeating a larger geographic region through various supply-chain connections (Ewing et al. 2005b).

Description of Hurricanes Rita and Ike

Hurricane Ike was the third most destructive hurricane to ever make landfall in the United States. It made final landfall on Saturday, Sept. 13, 2008, over Galveston, Texas, around 2 AM, as a Category 2 hurricane with maximum sustained winds nearing 110 mph (175 km/h) and central pressure of 950mbr, and covered over 425 miles of Texas coastline (NHC 2008). Harris County took a direct hit with the eye of the storm passing over Galveston Bay. After making landfall, Ike traveled north up Galveston Bay with a high storm surge along the east side of Houston (Berg 2009).

The Houston MSA, which includes Harris County and the City of Houston, the fourth largest city in the United States, had about 462,000 residential housing units. Approximately 2,900 units were deemed uninhabitable at the cost of over $208 million in damage. Around 251,000 residential housing units sustained minor damage from Hurricane Ike. Total damage was estimated up to $4.6 billion (Harris County, Texas 2009).

374 The American Journal of Economics and Sociology

About 2.6 million customers lost power in Texas and Louisiana. Downtown Houston experienced significant wind-induced damage as the city’s tallest building—the 75-story JP Morgan Chase skyscraper— had many windows broken (Clark 2008).

Ports from Corpus Christi to Lake Charles were shut down ahead of Hurricane Ike but it inflicted severe damage to the ports of Galveston and Houston and destroyed at least 10 offshore oil rigs and damaged several large pipelines. The U.S. Department of Energy reported that 14 oil refineries were closed by the storm, as well as two Texas strategic petroleum reserve sites (FEMA 2009), which caused the rise of gas prices and gas shortages across parts of the United States. Debris in Galveston Bay and the Houston ship channel postponed the reopening of these facilities for several days, leaving almost 150 tankers, cargo vessels, and container ships waiting offshore (EIA 2008).

Hurricane Rita reached its peak of Category 5 strength (on the Saffir-Simpson Hurricane Scale) along the central Gulf of Mexico. It made landfall between Sabine Pass, Texas, and Johnsons Bayou, Louisiana on September, 24, 2005 as a Category 3 storm on the Saffir-Simpson scale, with winds in excess of 120 miles per hour, and continued through southeastern Texas. The destruction was enormous, with the cost of reconstruction up to $12 billion (Knabb et al. 2006).

The town of Cameron took a direct hit, while high wind and flooding caused heavy damage to coastal communities from Port Arthur, Texas all the way to Terrebonne Parish. Approximately 2,000 square miles of farmland and marshes were inundated with sea water, producing indeterminate damage to the soil and the environment (Kurth and Burckel 2006). Many highways and minor roads were impassable after the hurricane near the landfall at the Louisiana and Texas border, and it was impossible to reopen quickly. Many residents were trapped for over 10 hours in traffic jams as a result of the massive evacuation order (Gibeaut et al. 2008).

Two refineries in Port Arthur, Texas were damaged by Hurricane Rita. Oil platforms and drilling rigs in its path were shut down and their workers evacuated, halting 98 percent of oil and natural gas production in the gulf (Kumins and Bamberger 2006). Houston escaped major damage from Hurricane Rita, some windows were

Economic Resiliency to Hurricanes 375

broken in downtown skyscrapers, and the storm downed some trees and signals. Rita missed the oil-refining region near Houston and Galveston.

Data for the Houston Metro Area

Monthly employment data for the Houston metropolitan statistical area (MSA) are obtained from the Bureau of Labor Statistics (http:// www.bls.gov) and the number of new privately-owned housing units authorized by building permits are retrieved from the U.S. Census Bureau (http://www.census.gov). Retail prices for all grades of gaso- line, including taxes paid by the consumer, are provided by the Energy Information Agency for the Houston MSA, which begins in June 2000 (http://www.eia.gov). Thus, our sample period is from June 2000 through October 2011. Following convention, we begin by seasonally adjusting the data series using the Census X-12 methodology. The gasoline price is later converted to a real price measure using the consumer price index. The natural log is taken of each variable and used in the analyses that follow.2

Table 1 provides descriptive statistics for the three variables used in this study. E denotes the log of the employment, G denotes the log of the real retail gasoline price, and P denotes the log of new building permits. Note that the working sample actually begins in June 2001 due to the monthly seasonal adjustment process. Figures 1–3 show plots of the three variables in levels. Perhaps somewhat strikingly, casual observation of the plots of these three variables reveals virtu- ally nothing about the impact of Hurricanes Rita and Ike. In fact, nothing more than a blip can be detected in the month of or the month immediately following these two hurricanes. Of course, this does not imply that there is not economic or engineering-related impact of these hurricanes. In fact, quite the opposite may actually be the case, as the three variables may be capturing broader measures of economic activity and the engineering environment that may only be revealed upon a closer look at their underlying, fundamental, and interactive relationship. Thus, we proceed to a formal investigation of the time series properties of these variables and a dynamic mod- eling technique known as vector autoregression, an error correction

376 The American Journal of Economics and Sociology

modeling that will allow us to understand more about the resiliency of the Houston area.

Methodology

Univariate Analysis and Stationarity Tests

It is important to consider the univariate properties of the variables under investigation in order to determine the proper specification of a vector autoregression (VAR) model. In standard VAR modeling, it is appropriate to use a stationary series. If the series has a unit root, it will be necessary to first-difference the series to achieve time-invariant linear properties and render a stationary process. Further, if two or more nonstationary series are each integrated of order one, I(1), it is possible that a linear combination of them is stationary, I(0). In this

Table 1

Descriptive Statistics

Panel A: Levels of the variables

E G P

Mean 14.716 0.0476 7.9251 Max. 14.8239 0.5089 8.6433 Min. 14.6355 −0.4565 7.0890 Std. Deviation 0.0532 0.2578 0.3296

Panel B: First-differences of variables

ΔE ΔG ΔP

Mean 0.0014 0.0040 7.00x10−5

Max. 0.013 0.1950 0.3272 Min. −0.0076 −0.3585 −0.2483 Std. Deviation. 0.0026 0.0646 0.0916

Notes: E denotes the log of the employment, G denotes the log of the real retail gasoline price, and P denotes the log of new building permits. Δ denotes first-difference operator. Adjusted sample period is June 2001–October 2011.

Economic Resiliency to Hurricanes 377

case, it is appropriate to use an error correction model (VAR model that allows for cointegration). In the case of no cointegration an unconstrained VAR will be appropriate.

Two separate unit root tests are performed to examine the stationarity of the respective time series. The augmented Dickey-Fuller (ADF) test is based on the estimation of the following equation (Dickey and Fuller 1981):

Δ ΔX t X Xt t i t i t i

N = + + − + +− −=∑κ δ ρ ψ υ( )1 1 1 (1)

where Xt is the individual series under investigation, Δ is the first- difference operator, t is a linear time trend, υt is a covariance stationary random error, and the number of lags on the augmenting term, N, is determined by Akaike’s information criterion to ensure serially uncorrelated residuals. The null hypothesis that X is a nonstationary series is rejected if (ρ − 1) < 1 and statistically significant.

Phillips and Perron (1988) developed an alternative unit root test that allows for weak dependence and heterogeneity in the error term.

Figure 1

Natural Log of Employment

14.60

14.64

14.68

14.72

14.76

14.80

14.84

2000 2002 2004 2006 2008 2010

E

378 The American Journal of Economics and Sociology

The test is robust to a wide range of serial correlation and time- dependent heteroscedasticity. The Phillips and Perron (PP) test for a unit root is based on the following regression:

X t T Xt t t= + − + +−ξ ξ λ γ0 1 12( ) (2)

where (t − T/2) is the time trend, with T representing the sample size, and γt is the error term. The null hypothesis of a unit root, Ho: λ = 1, is tested against the alternative hypothesis that Xt is stationary around a deterministic trend (Ha: λ < 1). As in the ADF test, MacKinnon (1991) critical values may be used to determine statistical significance for the Phillips-Perron test.

The results of the ADF and PP unit root tests are presented in Table 2 and suggest that each series is nonstationary in levels. A shock to one of these series is permanent as the series will not return to its long-run mean value over time. However, the first-difference of each series is found to be stationary. As mentioned above, it is possible for

Figure 2

Natural Log of Real Retail Gasoline Prices

-.6

-.4

-.2

.0

.2

.4

.6

2000 2002 2004 2006 2008 2010

G

Economic Resiliency to Hurricanes 379

a linear combination of nonstationary I(1) variables to be stationary. In this case, the variables are said to be cointegrated and while each may individually have no tendency to return to a particular value, there exists an attractor of which the group (of two or possibly more variables) tends toward. Given the results of the ADF and PP tests, we proceed to examine the cointegrating properties of the variables.

Given the multivariate nature of our model, we use the maximum likelihood method of Johansen (1988) and Johansen and Juselius (1990) to determine the number of cointegrating vectors. The pro- cedure is based on the vector autoregression (VAR) model of order p:

Δ Γ Δ Γ Δ Π

Γ Π Π

Π

X X X X

where

I

and

I

t t k t k t k t

i i

= + + + − +

= − + + +

=

− − − −μ ε1 1 1

1

…

…

−− − −Π Π1 … k

(3)

Figure 3

Natural Log of New Building Permits

6.8

7.2

7.6

8.0

8.4

8.8

2000 2002 2004 2006 2008 2010

P

380 The American Journal of Economics and Sociology

Xt is the vector of nonstationary I(1) variables under investigation and εt are independent Gaussian variables in k dimensions with mean zero and variance Ω. The vector Γi contains the short-run parameters that capture the disequilibrium feature of the data. The Π (k × k) matrix contains information about the long-run relationship that may exist among the three variables under investigation. The cointegration test determines the rank of the matrix Π. There are three possible cases to consider: (i) if r = 3, then the matrix Π has full rank, Xt is stationary in levels, and a traditional VAR model is appropriate; (ii) if r = 0 and Π has zero rank, then Xt is I(1) and the variables are not cointegrated; in this case, a traditional VAR in the first-differences is appropriate; and (iii) if the rank of Π is 0 < r < 3, then there are r cointegrating vectors and (p − r) stochastic trends. In this case, the error correction model is appropriate.

Two tests are conducted in order to determine the rank of Π, the trace test and the maximum eigenvalue test. The likelihood ratio statistic for (p = r) is specified as:

Q Tr i r

p

= − − +

∑ ln( )1 1

λ

Table 2

Unit Root Tests

ADF test statistic PP test statistic

Levels E −1.5525 −1.6984 G −2.7099 −2.7973 P −1.8336 −1.9407

First-differences ΔE −13.2090* −13.1332* ΔG −8.8610* −7.7143* ΔP −15.2038* −15.3506*

Notes: E denotes the log of the employment, G denotes the log of the real retail gasoline price, and P denotes the log of new building permits. Δ denotes first-difference operator. ADF denotes augmented Dickey-Fuller and PP denotes Phillips and Perron. A *denotes statistically significant at the 1 percent level or less based on MacKinnon (1991) critical values.

Economic Resiliency to Hurricanes 381

where λi is the largest eigenvalue (characteristic root), T is the number of observations, and Qr is the trace statistic used to test the null hypothesis H1(r) that the cointegrating vectors are less than or equal to r against the alternative Ha(p). The maximum eigenvalue statistic is based on the null hypothesis that the number of cointegrating vectors is r against the alternative r + 1, and is given by:

Q Tr r rmax( , ) ln( )+ += − −1 11 λ

Johansen and Juselius (1990) and Osterwald-Lenum (1992) pro- vide critical values for both tests. Table 3 reports results for the cointegration tests and indicates that there is one cointegrating vector. Thus, in what follows, we estimate a vector autoregression error correction model (VECM).

We estimate the error correction model (VECM) model. By design, the VAR nature of the model focuses on how past changes in one variable affect the current value of another variable. Thus, one typi- cally interprets the VAR output as representing how expectations about some variable are formed. Consider the following vector error correction model (VECM) described by Pesaran and Shin (1998):

Δ Π Γ Δ ΠΛx x x w ut t i t i i

p

t t= − + + + =− − =

−

∑1 1

1

, t 1, 2, , T,… (4)

where Π Φ Γ Φ= − = − = = + ∑ ∑I m i i

p

i j

j i

p

1 1

, for i = 1, . . . p − 1, and Λ is an

m × g matrix of unknown coefficients.

Table 3

Johansen-Juselius Cointegration Test Results

Hypothesized number of cointegrating equations (r) Trace Max Eigenvalue

0 37.0213* 26.1205* 1 10.9008 10.4513 2 0.4495 0.4495

Note: *denotes rejection of the null hypothesis at the 5 percent significance level based on MacKinnon-Haug-Michelis (1999) critical values. Both tests indicate one cointegrating equation.

382 The American Journal of Economics and Sociology

A VAR error correction model is estimated where the three equa- tions of the system correspond to E, G, and P. The lag order of the VAR is six based on the Akaike’s information criterion, Schwartz Bayesian criterion, and likelihood ratio tests. The long-run error estimated from the levels of the variables, lagged one period, is entered as the error correction term (−Πxt−1). The coefficients on this term pro- vide information on the adjustment process from disequilibrium. The cointegrating relationship implies the series form a stable long-run equilibrium relationship. Results are discussed in the next section. Additionally, the model includes indicator variables equal to one in the month of landfall for Hurricanes Rita and Ike and equal to zero elsewhere. The use of the hurricane variables serves to measure and control for the possible impact (i.e., similar to traditional event study methodology).

Discussion of Results

A critical element in error correction modeling is that at least one of the error correction terms in the system be negative for the cointegrating relationship to be established. If this condition holds, then the signs and coefficients on any remaining error correction terms may be interpreted in the usual manner. In a multivariate cointegrated system, the error correction representation relates each variable’s growth rate to previous departures from the vector’s equi- librium. Significant coefficients on what is called the speed of adjust- ment variable move the vector toward equilibrium (Cutler, Davies, and Schmidt 2000). Table 4 presents a summary of the VECM results.

Table 4

Summary of Vector Error Correction Model (VECM) Results

ΔE ΔG ΔP

Error correction term −0.0081 0.8000 −0.8076 (−0.6446) (3.0328) (−1.9466)

Notes: t-statistics in parentheses. E denotes the log of the employment, G denotes the log of the real retail gasoline price, and P denotes the log of new building permits.

Economic Resiliency to Hurricanes 383

Several points are worth noting. First, the coefficient on the lagged value of the error correction term can be interpreted as the speed of adjustment toward eliminating a previous period’s disequilibrium in the long-run cointegrating relation. The results presented in the first row indicate that it is the engineering component (building permits) and the energy component (real gasoline price) that change in response to a deviation from the long-run equilibrium between the E, G, and P. This finding is consistent with businesses responding to disruptions by changing employment levels in the short run, but over time the economic system returns to its stable relationship with adjustments being made in built environment and energy usage.

In fact, when there is a positive deviation (long-run error > 0), building permits adjust downward (note the negative and significant coefficient). Alternatively, for the case of a negative deviation (long- run error < 0) as in the case of a hurricane, building permits adjust upward to eliminate the disequilibrium. Further, real retail gasoline prices adjust upward with a positive deviation (note the positive and significant coefficient) and downward with a negative deviation. Employment (labor market measure) appears not to adjust to long-run disequilibrium as the estimated coefficient on the error correction term is insignificant. In the case of a hurricane event, employment is negatively impacted in the short term (immediate period/month fol- lowing the event), which lends itself to a long-run negative error as actual employment is less than predicted in the absence of the hurricane. Our model indicates that over the longer run, the built environment will increase (as reconstruction takes place) with changes in building permits reducing the disequilibrium error by nearly 81 percent in the first month (see the estimated coefficient of −0.8076), referred to as the speed of adjustment. Note that a negative sign on the speed of adjustment coefficient indicates that the corre- sponding change in the variable of interest, in this case ΔP, moves in the opposite direction to eliminate the disequilbrium. For the case of hurricanes, growth in building permits initially drops due to the disruption but over the longer term must rise to bring the economy back to normal. The corresponding speed of adjustment estimate for long-run error reduction is approximately 80 percent with respect to the equation for the real price of gasoline with prices falling in real

384 The American Journal of Economics and Sociology

terms (thus giving a boost to economic development as costs asso- ciated with transportation and associated activities fall). The speed with which real gas prices and building permits adjust to eliminate disequilibria is quite similar, although in opposite directions. The corresponding “half-lives” of the adjustment processes are each around three months, that is, half of the disequilibrium is eliminated within about three months (computed as xt = 0.5, where x is the estimated speed of adjustment coefficient and t is the number of periods in months). Moreover, the estimated coefficients indicate that less than 8 percent of the disequilibrium remains after 12 months (0.8112 ≈ 0.08 and 0.8012 ≈ 0.07).

Though not reported, it is interesting to note that the hurricane variables were only significant (and negative) in the (change in) real gas price equation. Thus, in terms of modeling this engineering- economic system, the impact from the hurricanes is to lower real gas prices. This finding is actually consistent with shortages, as many stations are shut down or inaccessible, gasoline supply is disrupted via transportation problems, and overall prices for goods and services rise more than gasoline prices, which is consistent with demand surge and induced inflation relative to gas prices in the very short term. Finally, in order to examine short-run effects, we conducted traditional Granger-causality tests. The results are not reported here, but they indicate that employment (positively) Granger-caused gas prices and (negatively) permits, and gas prices (positively) Granger-caused permits. All other short-run impacts were insignificant. These findings are generally consistent with the notion that postdisaster employ- ment growth, often spurred by workers hired for reconstruction and cleanup, is due to outside workers who increase the demand for gas and utilize temporary housing. Further, increases in gas prices appear to go hand-in-hand with more permits, possibly due to greater usage of construction vehicles and transportation of building materials.

Concluding Remarks

The potential applications of the model being presented are two-fold. First, it provides a new method for assessing disaster resiliency, an elusive term that was previously conceptualized but seldom examined

Economic Resiliency to Hurricanes 385

and measured in a vigorous manner. For example, Bruneau et al. (2003) proposed to define seismic resiliency of communities in four dimensions—technical, organizational, societal, and economic. Emmer and Swann (2007) developed a self assessment tool to calcu- late the Resiliency Index that estimates the exposure of the community to a disaster. The index was set as low, medium, and high, based on yes/no answers to a series of questions on critical facilities, evacu- ation, disaster preparation, and mitigation. Liu et al. (2006) created an index of 37 economic and social indicators that measure the impact of rebuilding efforts in Orleans Parish, the New Orleans metropolitan area, Louisiana, and Mississippi. Our future plan is to include other hurricane-prone regions on the Atlantic and Gulf coasts to confirm the presence or absence of a cointegrating relation. The result would enable us to distinguish resilient communities from less resilient ones and probe the underlying reasons.

Second, this study could serve as a basis for evaluating the merit and effectiveness of various strategies that are currently employed by the federal government under the Stafford Act. Specifically, the Staf- ford Act makes funds available to state, tribal, and local governments and dictates the process that must be followed in order for areas to be eligible for such aid (Bazan 2005). A priority of policymakers might be to speed up spending on reconstruction and restoration of infrastruc- tures, as they contribute to long-term recovery and sustainability of a local community. Employment, on the other hand, tends to adjust upward when the overall economy improves. At the same time, there will be a regional variation in terms of how quickly a local economy would respond to government actions. Therefore, a custom- ized approach towards federal assistance in disaster recovery might be more desired and possible.

Notes

1. HAZUS MH is the Federal Emergency Management Agency’s (FEMA’s) software and methodology for estimating potential losses from disasters.

2. We actually examined a number of other variables to capture the engineering and transportation-energy factors, such as the price of West Texas Intermediate crude oil, home prices, and airline arrivals and depar- tures at Houston Hobby and Bush Intercontinental Airports. Additionally,

386 The American Journal of Economics and Sociology

we considered wind speed at Houston Hobby and Bush as a measure of atmospheric conditions. However, in keeping with our goal of developing a parsimonious model with extension capabilities, the three variables— employment, real gas price, and housing permits—were chosen based on various goodness-of-fit measures and on the fact that these three variables are readily available for most hurricane-prone metro areas.

References

American Society of Civil Engineers. (2007). Hurricane Katrina Damage Assessment. Ed. Stephen A. Curtis. Reston, VA: American Society of Civil Engineers.

Barron, J., B. T. Ewing, and G. Lynch. (2006). Understanding Macroeconomic Theory. New York: Routledge.

Bazan, E. (2005). Robert T. “Stafford Disaster Relief and Emergency Assistance Act: Legal Requirements for Federal and State Roles in Declarations of an Emergency or a Major Disaster.” Congressional Research Service Report for Congress, RL33090.

Berg, R. (2009). Tropical Cyclone Report Hurricane Ike. National Hurricane Center. http://www.nhc.noaa.gov/pdf/TCR-AL092008_Ike_3May10.pdf.

Boin, A., and A. McConnell. (2007). “Preparing for Critical Infrastructure Breakdowns: The Limits of Crisis Management and the Need for Resili- ence.” Journal of Contingencies and Crisis Management 15(1): 50–59.

Bruneau, M., S. Chang, R. Eguchi, G. Lee, and T. O’Rourke et al. (2003). “A Framework to Quantitatively Assess and Enhance the Seismic Resilience of Communities.” Earthquake Spectra 19(4): 733–752.

Chow, E., and J. Elkind. (2005). “Hurricane Katrina and US Energy Security.” Global Politics and Strategy 7(4): 145–160.

Clark, A. C. (2008). The Vision Hurricane IKE. Houston-Galveston Area Council Transportation Department. http://www.h-gac.com/taq/ publications/newsletters/vision_Oct08.pdf.

Comfort, L. K., and T. W. Haase. (2006). “The Impact of Hurricane Katrina on Communications Infrastructure.” Public Works Management Policy 10(4): 328–343.

Cutler, H., S. Davies, and M. Schmidt. (2000). “Forecasting in a Large Macro- economic System.” Applied Economics 32: 1711–1718.

Dahlhamer, J. M., and K. J. Tierney. (1996). Rebounding from Disruptive Events: Business Recovery Following the Northridge Earthquake. Newark, DE: Disaster Research Center, University of Delaware.

Deltas, G. (2008). “Retail Gasoline Price Dynamics and Local Market Power.” Journal of Industrial Economics 56: 613–628.

Dickey, D. A., and W. A. Fuller. (1981). “Likelihood Ratio Statistics for Auto- regressive Time Series with a Unit Root.” Econometrica 49(4): 1057–1071.

Economic Resiliency to Hurricanes 387

Emmer, R., and L. Swann. (2007). Resiliency Index: Community Self- Assessment. http://www.gulfofmexicoalliance.org/projects/files/56Coastal _Resilience_Index_training_tool.pdf.

Enders, W. (2009). Applied Econometric Time Series, 3rd edition. NewYork: John Wiley & Sons.

Energy Information Administration. (2008). EIA Report on Hurricane Impacts on U.S. Energy. http://www.eia.gov/oog/special/hurricanes/gustav.html.

Engle, R. F., and C. W. J. Granger. (1987). “Cointegration and Error Correc- tion: Representation, Estimation, and Testing.” Econometrica 55(2): 251– 276.

Ewing, B., and J. Kruse. (2002). “The Impact of Project Impact on the Wilmington, North Carolina Labor Market.” Public Finance Review 30: 296–309.

Ewing, B. T., S. Hein, and J. Kruse. (2006a). “The Response of Insurer Stock Prices to Natural Hazards.” Weather and Forecasting 21: 395–407.

——. (2006b). Banking and Natural Disasters. Report to the Federal Deposit Insurance Corporation. Arlington, VA: Center for Financial Research.

Ewing, B. T., J. Kruse, and M. Thompson. (2003). “A Comparison of Employ- ment Growth and Stability Before and After the Fort Worth Tornado.” Environmental Hazards 5: 83–91.

——. (2004). “Employment Dynamics and the Nashville Tornado.” Journal of Regional Analysis and Policy 34: 47–60.

——. (2005a). “An Empirical Examination of the Corpus Christi Unemploy- ment Rate and Hurricane Bret.” Natural Hazards Review 6: 191–196.

——. (2005b). “Transmission of Employment Shocks Before and After the Oklahoma City Tornado.” Environmental Hazards 6: 181–188.

——. (2009). “Twister! Employment Responses to the 3 May 1999 Oklahoma City Tornado.” Applied Economics 41: 691–702.

Ewing, B. T., J. B. Kruse, and Y. Wang. (2007). “Local Housing Price Index Analysis in Wind-Disaster-Prone Areas.” Natural Hazards 40: 463–483.

Federal Emergency Management Agency. (2006). Hurricane Katrina in the Gulf Coast: Mitigation Assessment Team Report, Building Performance Obser- vations, Recommendations, and Technical Guidance. FEMA 549. http:// www.fema.gov/media-library/assets/documents/4069?id=1857.

——. (2009). Hurricane Ike in Texas and Louisiana: Mitigation Assessment Team Report. FEMA-3539. http://www.fema.gov/library/viewRecord.do? id=3539.

Federal Trade Commission. (2006). Investigation of Gasoline Price Manipula- tion and Post-Katrina Gasoline Price Increases: A Commission Report to Congress. Washington, DC: Federal Trade Commission.

Gibeaut, J., T. Hepner, J. Andrews, R. Gutierrez, and B. Smyth. (2008). Hurricane Rita Impacts on the Texas Shoreline. Austin, TX: University of Texas at Austin.

388 The American Journal of Economics and Sociology

Greenberg, M., M. Lahr, and N. Mantell. (2007). “Understanding the Economic Costs and Benefits of Catastrophes and Their Aftermath: A Review and Suggestions for the US Federal Government.” Risk Analysis 27(1): 83–96.

Guimaraes, P., F. L. Hefner, and D. P. Woodward. (1993). “Wealth and Income Effects of Natural Disasters: An Econometric Analysis of Hurricane Hugo.” Review of Regional Studies 23: 39–53.

Harris County Texas. (2009). Harris County Damage Assessment. http:// www.harriscountytx.gov/cmpdocuments/103/ike/harris_county_damage _assessment_ike.pdf.

Hunt, S. D. (1995). “The Resource-Advantage Theory of Competition: Toward Explaining Productivity and Economic Growth.” Journal of Management Inquiry 4: 317–332.

Irish, J. L., D. T. Resio, and J. J. Ratcliff. (2008). “The Influence of Storm Size on Hurricane Surge.” American Meteorological Society 38(9): 2003–2013.

Jeffrey, A., and E. Anne. (2008). “The Effect of Hurricane Katrina on the Labor Market Outcomes of Evacuees.” American Economic Review 98(2): 43–48.

Johansen, S. (1988). “Statistical Analysis of Cointegration Vectors.” Journal of Economic Dynamics and Control 12(2–3): 231–254.

Johansen, S., and K. Juselius. (1990). “Maximum Likelihood and Inference on Cointegration with Applications to the Demand of Money.” Oxford Bulletin of Statistics 52: 169–210.

Kaisera, M. J., Y. Yua, and C. J. Jablonowski. (2009). “Modeling Lost Produc- tion from Destroyed Platforms in the 2004–2005 Gulf of Mexico Hurricane Seasons.” Energy 34(9): 1156–1171.

Knabb, R. D., D. P. Brown, and J. R. Rhome. (2006). Tropical Cyclone Report Hurricane Rita. Washington, DC: National Oceanic and Atmos- pheric Administration.

Kumins, L., and R. Bamberger. (2006). “Oil and Gas Disruption from Hurri- canes Katrina and Rita.” Congressional Research Service. The Library of Congress.

Kurth, M. M., and D. V. Burckel. (2006). The Rita Report: A Summary of the Social and Economic Impact and Recovery of Southwest Louisiana One Year after Hurricane Rita. Louisiana Recovery Authority.

Lewis, M. S. (2009). “Temporary Wholesale Gasoline Price Spikes Have Long-Lasting Retail Effects: The Aftermath of Hurricane Rita.” Journal of Law and Economics 52(3): 581–605.

Liang, D., A. Dabral, H. He, and R. Vega. (2005). Hurricane Season. Texas Tech University, Wind Science and Engineering Library.

Liu, A., M. Fellowes, and M. Mabanta. (2006). Katrina Index Tracking Vari- ables of Post-Katrina Recovery. Washington, DC: Brookings Institution, Special Analysis in Metropolitan Policy.

Lodree, E., and S. Taskin. (2009). “Supply Chain Planning for Hurri- cane Response with Wind Speed Information Updates.” Computers &

Economic Resiliency to Hurricanes 389

Operations Research: Special Issue on Disaster Recovery Planning 36: 2–15.

MacKinnon, J. G. (1991). “Critical Values for Cointegration Tests.” In Long- run Economic Relationships. Eds. R. F. Engle and C. W. J. Granger, pp. 266–276. Oxford: Oxford University Press.

Marshall, T. P. (2006). “Hurricane Katrina Damage Survey.” In 27th Conference on Hurricanes and Tropical Meteorology.

Marshall, C. C. (2008). Maintaining Personal Information: Issues Associated with Long Term Storage Preservation. http://www.csdl.tamu.edu/ ~marshall/PIM%20Chapter-Marshall.pdf.

McCarthy, K., D. J. Peterson, N. Sastry, and M. Pollard. (2006). The Repopulation of New Orleans, After Hurricane Katrina. Santa Monica, CA: RAND. http://www.csdl.tamu.edu/~marshall/PIM%20Chapter-Marshall. pdf.

Mosqueda, G., and K. A. Porter. (2007). Damage to Engineered Buildings and Bridges in the Wake of Hurricane Katrina. Structure Congress 2007 Forensic Engineering.

National Hurricane Center. (2008). Hurricane IKE Advisory Archive 48b. http://www.nhc.noaa.gov/archive/2008/al09/al092008.public_b.048. shtml?

National Institute of Standards and Technology. (2006). “Performance of Phy- sical Structures in Hurricane Katrina and Hurricane Rita: A Reconnais- sance Report.” NIST Technical Note 1476.

National Research Council. (2009). Strengthening Forensic Science in the United States: A Path Forward. http://judiciary.house.gov/hearings/ printers/111th/111-28_49681.PDF.

O’Connor, J. S., and P. E. McAnany. (2008). “Damage to Bridges from Wind, Storm Surge and Debris in the Wake of Hurricane Katrina.” MCEER Special Report Series.

Osterwald-Lenum, M. (1992). “A Note with Quantiles of the Asymptotic Distribution of the Maximum Likelihood Cointegration Rank Test Statis- tics.” Oxford Bulletin of Economics and Statistics 54: 461–471.

Phillips, P., and P. Perron. (1988). “Testing for a Unit Root in Time Series Regression.” Biometrika 75: 335–346.

Plyer, A., J. Bonaguro, and K. Hodges. (2010). “Using Administrative Data to Estimate Population Displacement and Resettlement following a Cata- strophic US Disaster.” Population and Environment 31: 150–175.

Saleem, K., S. Luis et al. (2008). “Towards a Business Continuity Information Network for Rapid Disaster Recovery.” In 2008 International Conference on Digital Government Research, pp. 107–116.

Stephen, P. L., G. C. Arindam, and J. Carolyn. (2007). “Wall of Wind Full-Scale Destructive Testing of Coastal Houses and Hurricane Damage Mitigation.” Journal of Coastal Research 24(1): 113–119.

390 The American Journal of Economics and Sociology

Tang, C. S. (2006). “Perspectives in Supply Chain Management: A Review.” International Journal of Production Economics 103: 451–488.

Taskin, S., and E. J. Lodree. (2010). “Inventory Decisions for Emergency Supplies Based on Hurricane Count Predictions.” International Journal of Productions Economics 126: 66–75.

Tatano, H., T. Homma, N. Okada, and S. Tsuchiya. (2004). “Economic Resto- ration After a Catastrophic Event: Heterogeneous Damage and Infrastruc- ture and Capital and its Effects on Economic Growth.” Journal of Natural Disaster Science 26(2): 81–85.

Tatum, M. C., and F. Terrell. (2012). “Hurricane Reconstruction in the United States Gulf Coast.” International Journal of Disaster Resilience in the Built Environment 3(3): 199–219.

Vigdor, J. (2008). “The Economic Aftermath of Hurricane Katrina.” Journal of Economic Perspectives 22(4): 135–154.

Webb, G. R., K. J. Tierney, and J. M. Dalhammer. (2000). “Businesses and Disasters: Empirical Patterns and Unanswered Questions.” Natural Hazards Review 1(2): 83–90.

Womble, J., S. Ghosh, C. Friedland, and B. Adams. (2006). “Hurricane Katrina—Advanced Damage Detection: Integrating Remote-Sensing Images with VIEWSTM Field Reconnaissance.” Report of the Multidisci- plinary Center for Earthquake Engineering Research, University at Buffalo. http://mceer.buffalo.edu/publications/Katrina/06SP02-web.pdf.

Zhang, D. L., and X. Wang. (2003). “Dependence of Hurricane Intensity and Structure on Vertical Resolution and Time-Step Size.” Advances in Atmospheric Sciences 20(5): 711–725.

Economic Resiliency to Hurricanes 391