Assessing flood impacts on the regional property markets in Queensland, Australia.
Akbar, Delwar ; Rolfe, John ; Small, Garrick 等
1. INTRODUCTION
The main driver behind the growth of Australian regional towns in
recent years, especially those in Queensland, has been the development
of resources such as minerals, coal and natural gas. Regional
communities in Queensland are the economic power house of the state;
contributing more than 80 per cent of goods and services exports, as
well as about 16 billion dollars each year to the state's economy
(Queensland Government, 2008). However many regional towns and local
governments areas in Queensland were flooded in 2011 with at least
seventy towns and over 200 000 people affected (Australian Government,
2014). Rockhampton is one of the regional cities within central
Queensland that was severely affected by flood in 2011, as it was
disconnected from the state's administrative and commercial capital
(Brisbane) by road, air and rail for more than one week. Rolfe et al.
(2013) have estimated that the direct economic impact of the flood on
Rockhampton and the associated highway and airport closure was
approximately $66.7 million.
There is potential for longer term impacts of floods on housing
markets, as residents re-assess flood risks and the personal and
economic disruption that they cause. Several studies have been conducted
to identify flood impacts on the property market in Australia (Eves and
Wilkinson, 2014; Small et al., 2013), finding minor and short term
impacts on property prices. Similar studies elsewhere (Bin and Polasky,
2004; Chou and Shih, 2001) have showed significant impacts of flood and
inundation events on the property market. However, none of these studies
have been able to identify which segment of the property market was most
affected.
This study focuses on identifying market vulnerability by comparing
segments of the property market i.e., number of total house (TH) sales
(old and new houses), new house and land (HL) package sales and land
only (LO) sales before and after the 2011 floods. This study also made
efforts to test whether flood impacts had been offset by the impact of
mining growth in this region.
The paper has been organised as follows: following this
introductory section, Section 2 provides contextual background for the
study; Section 3 gives a brief background of the study area; Section 4
describes data and methods; and Section 5 provides the findings and
analysis of the study. The paper concludes in section 6.
2. CONTEXTUAL BACKGROUND: FLOOD IMPACTS ON HOUSING
Floods have always had some level of impact on property markets
depending on their severity and inundation level (Worthington, 2008;
Troy and Romm, 2004). A number of studies have been conducted in the
USA, Germany, Taiwan and Australia to find out the effect of flooding on
local residential property markets (Eves and Wilkinson, 2014; Small et
al., 2013; Kropp, 2012; Bin et al., 2006; Bin and Polasky, 2004; Merz et
al., 2004; Chou and Shih, 2001). Most of these studies have found that a
flood event can decrease the value of inundated property or the
inundated part of the town, but not the overall property market at a
local or regional level (Figure 1). However, none of these studies have
estimated the effect of flooding on the local property market or what
segment of the property market is most affected.
The effects of flooding on property markets can be offset by the
effects of other local or regional factors such as resource
developments, regional population growth, new social facilities
development and increases in employment (Kropp, 2012). However, no one
has tested the effect of one or all of these factors on flood impacts on
property markets or whether impacts vary across subsets of those
markets.
The focus of this study is to test whether there are any
differences in flood impacts between the property submarkets at a local
level. Subsequently the study tests whether any local or regional
economic determinants can offset or enhance the flood impacts on housing
markets at a local level. Testing these hypotheses is very important in
predicting future housing markets as well as providing empirical
evidence to support policy makers in deciding what measures should be
undertaken in the aftermath of flooding. This type of research is
becoming essential for community and property investors because major
floods in this area previously occurred approximately every fifty years
but now appear to be happening at 10 to 25 year intervals (Kropp, 2012;
BOM, 2013).
[FIGURE 1 OMITTED]
3. STUDY AREA
This study focuses on Rockhampton, which is a regional city near
the mouth of the Fitzroy River in central Queensland (Figure 2). The
Fitzroy is the second largest externally draining catchment in
Australia, and experiences large floods in some seasons, including 2011,
with subsequent impacts on Rockhampton. There are advantages in using
Rockhampton as a case study; the city is large enough to generate
substantial property data, there are some areas of housing that are
affected by floods while others are not, and the use of a single city
minimises the impact of other confounding factors.
[FIGURE 2 OMITTED]
Figure 3 indicates major and minor flooding events that have
occurred
in Rockhampton including five major inundations since 1890. Small et
al. (2013) examined resident opinions in comparison to market realities
of the impact of flooding on property value. They found that over 50 per
cent of respondents believed that the flood event had a negative impact
on property values causing a decrease in values. Despite the beliefs of
respondents, a weak relationship between floods and the dynamics of
property markets was found. However, the reason for such weak
relationship was not explored. In contrast, CCIQ (2011) found evidence
of minor to major impacts of flooding in Queensland businesses within
the flood affected towns, including Rockhampton.
[FIGURE 3 OMITTED]
Rockhampton is an administration, service and population hub in
central Queensland. Its economy has been growing strongly since 2003
because of the mining boom in the nearby Bowen Basin region (Akbar et
al., 2010). In addition, large scale natural gas and infrastructure
development projects in the nearby port city of Gladstone contributed to
an increase in the resident population in Rockhampton, as Gladstone had
been suffering with housing availability and affordability difficulties
between 2009 and 2013 (Akbar et al., 2013). These growth pressures in
the regional economy may have offset any negative impacts of the 2011
flood event.
4. METHODS
A number of studies used qualitative, quantitative or mixed
methodology to identify the impacts of flooding on property market
(Table 1). However this study used a quantitative methodology with
longitudinal data of house sales, inundation levels and mining impacts
to answer the two research questions. Longitudinal data of property
sales in three segments of property markets in Rockhampton Regional
Council (i.e., a local government area in central Queensland region)
were collected from the Queensland Treasury and Trade (QTT) database on
residential land development activity profile (QTT, 2014; 2008). This
longitudinal data includes quarterly median price and number of sales
between quarter 1, 2000 and quarter 4, 2014. Flood inundation level data
was collected from the Australian Bureau of Meteorology (BOM). We used
dummy variables (i.e., no impact = 0, having impact = 1) for flood and
mining impacts between quarter 1, 2000 and quarter 4, 2014.
This study used independent t-Test with a 95 per cent confidence
level to identify the significant difference between property price and
sales before and after the 2011 flood in Rockhampton. We used the SPSS
Package-PASW Statistics 22 to do Independent Samples t-Tests.
Subsequently this study used multivariate regression models to
establish the relationship among the number of sales, median property
price, flood and mining impacts on the property markets. We used JMP-Pro
software for applying regression models and visualising the flood and
mining leverages on the number of property sales. The formula and the
findings from these models are described in the next section.
5. FINDINGS AND ANALYSIS
The January 2011 flood, with a peak water height of 9.2m, was the
most severe and devastating flood in the Rockhampton region over the
last twenty years. However, there were three other floods within this
period in 2008, 2010 and 2013, with flood peaks vary between 7.1m and
8.6m (Figure 3). The 2011 flood was chosen as a market intervention
point because of its severity. Data for quarterly median property price
and number of sales of all three segments of the property market are
only available between 2000 and 2014 from the Queensland
Government's published source (i.e, QTT, 2014 and 2008). Therefore
we used quarter 1, 2000 to quarter 4, 2010 property price and number of
sales data as before flood data and quarter 1, 2011 to quarter 4, 2014
as after flood data. Then we carried out independent t-tests to examine
whether any significant differences existed between property median
price and number of sales before and property median price and number of
sales after the 2011 flood's for each segment of the property
market. Here, the null hypothesis and the alternative hypothesis (termed
as [H.sub.o] and [H.sub.a] respectively) are as follows:
[H.sub.o] = There is no significant difference between before and
after the 2011 flood's property median price or number of sales of
a particular segment of the property market, and
[H.sub.a] = There is a significant difference between before and
after the 2011 flood's property median price or number of sales of
a particular segment of the property market.
The decision rule is given by: if p [less than or equal to]
[alpha], then reject [H.sub.o].
Considering the empirical results presented in Table 2, the
condition is satisfied for total house (TH), and new house and land (HL)
package sales (both for the median price and the number of sales
variables), but not the number of land only (LO) sales (Table 2). For
both total house sales and new house and land package sales, the
independent samples t-test had a significance level between 0.000 and
0.001, which is less than a (0.05) (Table 2). So these primary results
exhibited that flood affected significantly two segments of the property
market (i.e., TH sales and new HL packages sales). In contrast some
studies such as Small et al., 2013 and Eves and Wilkinson, 2014 reported
a very minimal or short term impact of flood on property markets in
flood affected areas of Australian cities. To further explore this
contradiction, this study tested property sales with property price,
flood and mining impact variables, to conclude the magnitude and
direction of the effects on the property market and determine the
flooding and mining leverages on the property markets.
Multivariate regression models (MRM) were used to add flood and
mining impacts in different quarters between 2000 and 2014. The median
value for houses in a quarter of a year (for example, quarter 1, 2000)
were regressed against the number of house sales in that particular
quarter. Essentially each quarter becomes an "observation" and
therefore we had 60 observations. Our generic formula to regress the
number of house sales in a quarter of a year is:
Y = [[beta].sub.0] + [[beta].sub.l][X.sub.l] + u (1)
Where Y is the number of predicted sales, [X.sub.1] is the median
sales prices, [[beta].sub.0] is the constant or intercept term,
[[beta].sub.1] is slope parameter and u is an unobserved random
variable, known as the error or disturbance term.
As there are two more predictors in equation 1 to capture flood and
mining impacts, our general MRM equation for predicting the number of
sales is thus:
Y = [[beta].sub.0] + [[beta].sub.1][x.sub.1] +
[[beta].sub.2][X.sub.2] + u (2)
Here [X.sub.1] is median price for "total house (TH)
sales" or "new house and land (HL) package sales" or
"land only (LO) sales" and [X.sub.2] represents either flood
impact or mining impact, and the following six equations predict the
number of sales of each market segment either considering flood or
mining impact (Equations 3-8).
Number of total house sales (TH) = 969.114 - 0.002 * Median Sale
Price (TH) -72.766 * Flood impact (3)
Number of new house and land package sales (HL) = 34.588 -0.0392 *
Median Sale Price (HL) - 1.218 * Flood impact (4)
Number of land only sales (LO) = 193.485 - 0.001 * Median Sale
Price (LO) - 7.108 * Flood impact (5)
Number of sales (TH) = 909.558 - 0.002 * Median Sale Price (TH) +
284.667 * Mining impact (6)
Number of new house and land package sales (HL) = 34.163 -0.0436 *
Median Sale Price (HL) + 3.216 * Mining impact (7)
Number of sales (LO) = 178.497 - 0.001 * Median Sale Price (LO) +
111.378 * Mining impact (8)
Flood usually affects the property market negatively (Lamond and
Proverbs, 2006) and mining affects the property market positively (Akbar
et al, 2013) in terms of number of property sales and median property
price in regional towns. Therefore we either considered flood impact or
mining impact at one time in each equation. Quarter 1, 2003 to Quarter
4, 2007 and Quarter 1, 2010 to Quarter 4, 2013 were considered as having
mining impact (dummy variable- having impact=1) on the Rockhampton
property market. On the other hand, the global financial crisis hit the
mining sector between quarter 1, 2008 and quarter 4, 2009 and the price
of coal started falling from quarter 1, 2014, so these years were
considered as no impact (i.e., 0) of mining along with the initial
period (quarter 1, 2000 to quarter 4, 2002) of not significant mining
activities in this region.
Based on these equations, an effect summary of the models is
provided in Table 3.
Here in Table 3 parameter estimates refer to magnitude and
direction of that relation i.e., mining or flooding impact with the
number of sales; the p-value is the result of the test of the following
null hypothesis, a p-value less than 0.05 means that the coefficient is
statistically significant. However, for a large or very significant
effect, the associated p-values are often very small and it is hard to
visualize these small values graphically. When transformed to the
LogWorth (-log10(p-value)) scale, highly significant p-values have large
LogWorths and non-significant p-values have low LogWorths. A LogWorth of
zero corresponds to a nonsignificant p-value of 1. Any LogWorth above 2
corresponds to a p-value below 0.01 and so on (JMP, 2015). Rsq value
represents the proportion of variation in the dependent variable (i.e.,
number of sales).
We need to consider all four values (parameter estimate, p-value,
LogWorth value and Rsq value) (Table 3) to find out the significance and
direction of the flood or mining impacts on the property market. Table 3
exhibited that mining had very significant and positive impact on the
number of total house (TH) sales in Rockhampton compare to a moderate
impact on land only sales and little or no effect on new house and land
(HL) package sales. On the other hand, flooding had moderately
significant but negative impact on the number of total house (TH) sales
and almost no or little impact on new house and land (HL) package sales
and land only (LO) sales.
Therefore our multivariate regression models indicate that both
mining and flooding affected the number of total house (TH) sales
significantly but positively and negatively respectively. We are yet to
understand how these factors (i.e., mining and flooding) leverage the
total house sales. The coefficient of each predictor variable is the
effect of that variable, which is termed as the leverage of each impact
i.e., either flood or mining. Here, the effect leverage plot for x2
(flood or mining impact) is essentially a scatterplot of the X-residuals
against the Y-residuals (Figures 4a and b).
[FIGURE 4a OMITTED]
[FIGURE 4b OMITTED]
These leverage plots (Figures 4a and b) show the impact of adding
mining or flood effects to the models at a 5% level of significance. In
addition, a visual indication of a significant effect of both mining and
flooding is the fitted regression line (i.e., red solid line). As the
mining impact leverage points are more horizontally distant from the
centre of the plot compared to that of flood impact, it shows that the
mining effects exert more influence than the flood effects. Both
interpretations (i.e., numeric LogWorth values (Table 3) and visual
observation of the leverage points (Figures 4a and b)-exhibited that
multiple regression models observed statistically significant impacts of
mining and flood on the number of total house sales but in opposite
directions. Growth in the mining sector had positive impacts on the
number of sales during this period and also offset the flood impact on
sales.
6. CONCLUSION
Flooding appears to have had an impact on the number of total house
sales in the property market of Rockhampton. Eves and Wilkinson (2014)
investigated the house price impact of the same flood in Brisbane and
they came to similar conclusions in terms of changes in median house
price over a short term. They did not specify the reasons for the low
impact of flooding on the property prices which suggests that a separate
study on the metropolitan market may be useful given its distinctive
economic bases and demographic characteristics. Earlier, Bin and Polasky
(2006) found a very significant value loss in the housing market after
Hurricane Floyd in September 1999 in North Carolina and again their
findings are almost the opposite to the primary findings of this study.
The primary finding in this study is that flooding has had a
significant effect on the number of total house (TH) sales in contrast
to the other two segments (i.e., new house and land packages sales and
land only sales) which were not significantly affected. By contrast,
this study differs from others (CCIQ, 2011; Lamond and Proverbs, 2006)
in its finding that the floods did not have a significant impact on
house prices.
It has also been found that the impacts of flooding on housing
markets in Rockhampton had been offset by mining impacts which explains
why the devastating 2011 flood did not significantly affect new house
and land package sales and land only sales. Also the mining impact on
total house sales is very significant and in a positive direction but
the flood impact on this market segment is low to moderately significant
and in a negative direction.
A single test or model cannot explain the impact of flooding on
regional property markets in Australian regional cities. However, a
method containing several models and effect assessments such as the one
used in this study can explain the flood impact more rigorously. Both
the method and the models can be used for further study. A lesson that
can be taken from this study is that local or regional development
factors can offset any flood impacts on housing markets either partially
or significantly at local and sub-regional scales. Therefore the policy
makers should emphasise enhancing the pre-existing development projects
through providing monetary or infrastructure support that can help the
local economy and help maintain housing markets.
ACKNOWLEDGEMENT: The research involved in this paper was funded by
CQUniversity Australia
REFERENCES
Akbar, D., Ian, C. and Rolfe, J. (2010). Impact of GFC in Housing
Market in Regional Australia: Lessons from the Central Queensland.
Proceedings of the APNHR Conference, Beijing, 2122 August.
Akbar, D., Rolfe, J. and Kabir, Z. (2013). Predicting Impacts of
Major Projects on Housing Prices in Resource Based Towns with a Case
Study Application to Gladstone, Australia. Resources Policy, 38, pp.
481-489.
Australian Government (2014). Brisbane and Queensland, 2010-2011
Floods. Natural Disasters in Australia. Online version accessed 28
October 2014, http://australia.gov.au/about-australia/australian-story/natural-disasters.
Bin, O. and Polasky, S. (2004). Effects of Flood Hazards on
Property Values: Evidence Before and After Hurricane Floyd. Land
Economics, 80(4), pp. 490-500.
Bin, O., Crawford, T., Kruse, J. and Landry, C. (2006). Flood Prone
with a View: Coastal Housing Market Response to Risk and Amenity. East
Carolina University.
BOM (2013). Water Level Data--Fitzroy River. Rockhampton: Bureau of
Meteorology, Australia.
Chamber of Commerce and Industry Queensland (CCIQ) (2011). Six
Months on from Queensland's Natural Disasters: A Report to the
Queensland Government. Brisbane: Chamber of Commerce and Industry
Queensland.
Chou, S.H. and Shih, H.C. (2001). Metropolis: Impact Assessment of
Flood Risk on Housing Property Market in Taipei. Proceedings of the 8th
European Real Estate Society Conference. ERES: Conference. Alicante,
Spain, 2001.
Eves, C. (2004). The Impact of Flooding on Residential Property
Buyer Behaviour: an England and Australian Comparison of Flood Affected
Property. Structural Survey, 22(2), pp. 84-94.
Eves, C. and Wilkinson, S. (2014). Assessing the Immediate and
ShortTerm Impact of Flooding on Residential Property Participant
Behaviour. Natural Hazards, 71, pp. 1519-1536.
JMP (2015). JMP Statistical Discovery from SAS, SAS, North
Carolina. Online version accessed 14 July 2015,
http://www.imp.com/support/help/The Response Screening Repo rt.shtml,
Kropp, S. (2012). The Influence of Flooding on the Value of Real
Estate. FIG Working Week 2012, Knowing to Manage the Territory, Protect
the Environment, Evaluate the Cultural Heritage, Rome, Italy, May 6-10.
Lamond, J. and Proverbs, D. (2006). Does the Price Impact of
Flooding Fade Away? Structural Survey, 24(5), pp. 363-377.
Merz, B., Kreibich, H., Thieken, A. and Schmidtke, R. (2004).
Estimation Uncertainty of Direct Monetary Flood Damage to Buildings.
Natural Hazards and Earth System Sciences, 4, pp. 53-163.
Pryce, G., Chen, Y. and Galster, G. (2011). The Impact of Floods on
House Prices: An Imperfect Information Approach with Myopia and Amnesia.
Housing Studies, 26(2), pp. 259-279.
Queensland Treasury and Trade (QTT) (2008). Residential Land
Activity Fact Sheet, December Quarter 2008. Brisbane: Queensland
Treasury and Trade.
Queensland Treasury and Trade (QTT) (2014). Residential Land
Development Activity Profile, March Quarter 2014. Brisbane: Queensland
Treasury and Trade.
Queensland Government (2008). Sustainable Resource Communities
Policy--Social Impact Assessment in the Mining and Petroleum Industries.
Brisbane: Department of Tourism, Regional Development and Industry.
Rolfe, J., Kinnear, S. and Gowen, R. (2013). Simplified assessment
of the regional economic impacts of interruption to transport corridors
with application to the 2011 Queensland floods. Australian Journal of
Regional Studies, 19(2), pp. 215-238.
Small, G., Newby, L. and Clarkson, I. (2013). Opinion Versus
Reality: Flood-affected Property Values in Rockhampton, Australia.
Proceedings of the Pacific Rim Real Estate Society international
conference 2013, PRRES: RMIT, Melbourne.
Troy, A. and Romm, J. (2004). Assessing the Price Effects of Flood
Hazard Disclosure Under the California Natural Hazard Disclosure Law
(AB1195). Journal of Environmental Planning and Management, 47(1), pp.
137-162.
Worthington, A.C. (2008). The Impact of Natural Events and
Disasters on the Australian Stock Market: A GARCH-M analysis of storms,
Floods, Cyclones, Earthquakes and Bushfires. Global Business and
Economics Review, 10(1), pp. 1-10.
Delwar Akbar
Senior PD Research Fellow, School of Business and Law, CQ
University, North Rockhampton, QLD, 4702, Australia. Email:
d.akbar@cqu.edu.au.
John Rolfe
Professor, School of Business and Law, CQ University, North
Rockhampton, QLD, 4702, Australia. Email: i.rolfe@cqu.edu.au.
Garrick Small
Associate Professor, School of Business and Law, CQ University,
North Rockhampton, QLD, 4702, Australia. Email: g.small@cqu.edu.au.
Rahat Hossain
Post Doctoral Research Fellow, School of Business and Law, CQ
University, Rockhampton, QLD, 4702, Australia. Email:
m.hossain@cqu.edu.au.
Table 1. Methods Used in Flood Impacts on Housing Market Studies.
Author Study Area Methods/Techniques/data source
and year
Eves and 48 suburbs out Weekly change analysis in sales,
Wilkinson, of 190 suburbs rental listings and volumes
2014 in Brisbane, of sales over a 12 month period
Queensland, Using RP database for housing
Australia data and Australian Bureau of
Statistics for socio-demographic
data
Small et al, Rockhampton, Mixed methods: Longitudinal
2013 Queensland median sales price sourced from
RP Database and a survey on the
flood affected families
Pryce et al, Theoretical Using a framework for modelling
2011 construct the housing market response to
flood impact considering g
myopic and amnesiac risks.
Using contemporary theories
and empirical evidences.
Lamond and Barby, North Case study with comparative
Proverbs, Yorkshire, UK analysis and applying semi-
2006 logarithmic regression model.
Using longitudinal property sell
and flood data from the Land
Registry and Environmental Agency
Bin and Pitt Country, Hedonic property price function
Pollasky, North Carolina, and comparison within and outside
2004 USA the floodplain area
Using Pitt Country's Geographic
Information Systems (GIS) and
Management Information Systems
(MIS) for distance, property
parcel records and property's
physical attributes data.
Eves, 2004 England and Survey method and data analysed
Wales, UK and by descriptive statistics
Sydney, Using questionnaire survey
Australia and property sales data
Table 2. Independent Samples t-Test: Results for Median Price and
Number of Property Sales.
Property sub- Test variable t-test for
market equality of Means
(equal variance assumed)
df Sig. (2-tailed)
Total house sales Median price 58 0.000
Number of sales 58 0.000
New house and land Median price 58 0.001
package sales Number of sales 58 0.001
Land only sales Median price 58 0.000
Number of sales 58 0.086
Source: the Authors
Table 3. Effect Summary of the Models.
Mining impact
Impact Total New Land
indicator house house only
sales and land sales
package
sales
Parameter 284.667 3.215 111.378
estimates
P-Value 0.000 0.249 0.000
LogWorth 6.581 0.604 6.554
Rsq value 0.559 0.271 0.401
Flood impact
Impact Total New Land
indicator house house only
sales and sales
land
package
sales
Parameter -72.766 -1.218 -
estimates 7.108
P-Value 0.345 0.703 0.804
LogWorth 0.462 0.153 0.095
Rsq value 0.306 0.256 0.046
Source: the Authors
