# RECORD 85ea3296a5d9175ea12142cf209523d5db6d0b42 # TITLE: Hedonic Analysis of the Impact of Flood Events on Residential Property Values in Malaysia: A Study of Willingness to Pay # YEAR: 2019.0 VENUE: Malaysian Journal of Economic Studies DOI: 10.22452/mjes.vol56no1.4 # AUTHORS: N. Ismail; M. Z. Abd Karim; B. Hasan-Basri # PAGES: 22 CANDIDATE TABLE PAGES (1-based): [11, 13, 15] # ABSTRACT: The effect of flooding on residential property values (RPV) is a major concern to all property buyers and owners. Although numerous studies have already discussed the risk of devaluation of those properties situated in flood prone areas, those that focused on the impact of flooding on residential properties in Malaysia are still limited. This paper extends existing literature by also estimating consumers’ willingness to pay for flood control measures to reduce the flood risk in Malaysia. Using the hedonic pricing model (HPM), our results suggest that the market value of urban and rural residential properties significantly decreases by 18.5 percent and 13.6 percent due to flooding, respectively. The result also shows that, respondents who are risk averse are willing to pay 35.4 percent more for flood control measures to reduce impact of flood risk compared to those who are risk takers. The results of this study can help property owners to understand the factors that contribute to property devaluation due to flooding. This study also proposes flood insurance programmes to be implemented as flooding is a major concern to real property owners. ===== PAGE 1 (score 5, numeric-paren 0, table-words 0, var-words 5) ===== Malaysian Journal of Economic Studies Vol. 56 No. 1, 2019 63 Malaysian Journal of Economic Studies 56(1): 63–84, 2019 ISSN 1511-4554 Hedonic Analysis of the Impact of Flood Events on Residential Property Values in Malaysia: A Study of Willingness to Pay Nur Hafizah Ismaila Mohd Zaini Abd Karimb Bakti Hasan-Basric Universiti Utara Malaysia Abstract: The effect of flooding on residential property values (RPV) is a major concern to all property buyers and owners. Although numerous studies have already discussed the risk of devaluation of those properties situated in flood prone areas, those that focused on the impact of flooding on residential properties in Malaysia are still limited. This paper extends existing literature by also estimating consumers’ willingness to pay for flood control measures to reduce the flood risk in Malaysia. Using the hedonic pricing model (HPM), our results suggest that the market value of urban and rural residential properties significantly decreases by 18.5 percent and 13.6 percent due to flooding, respectively. The result also shows that, respondents who are risk averse are willing to pay 35.4 percent more for flood control measures to reduce impact of flood risk compared to those who are risk takers. The results of this study can help property owners to understand the factors that contribute to property devaluation due to flooding. This study also proposes flood insurance programmes to be implemented as flooding is a major concern to real property owners. Keywords: Floods, hedonic price model, property, valuation JEL classification: Q54, R21, R32 1. Introduction Flood is a natural phenomenon that can have disastrous effect on the environment and on those affected communities (IFRC, 2009; Jha, Bloch, & Lamond 2012). It is also one of the most frequent and most costly due to the damages it can cause to properties, livelihoods and human health. In terms of asset loss, flooding can destroy public infrastructures, houses and other real and personal properties. The destruction that floods bring disrupts people’s daily lives, with some even forced to leave their homes. a School of Economics, Finance and Banking, Universiti Utara Malaysia, 06010 Sintok, Kedah, Malaysia. Email: hafizah19@live.com, n.hafizah.mohammad@uum.edu.my (Corresponding author) b Othman Yeop Abdullah Graduate School of Business, Universiti Utara Malaysia, 06010 Sintok, Kedah, Malaysia. Email: zaini500@uum.edu.my c School of Economics, Finance and Banking, Universiti Utara Malaysia, 06010 Sintok, Kedah, Malaysia. Email: bakti@uum.edu.my * The authors wish to thank the Ministry of Higher Education Malaysia for funding this study under the Long-Term Research Grant Scheme (Code LRGS/b-u/2012/UUM/Teknologi Komunikasi dan Informasi). The views expressed in this study are those of the authors and do not necessarily reflect the views or policies of the ministry or the project team. Any errors are the sole responsibility of the authors. Article Info: Received 26 June 2018; Revised 3 February 2019; Accepted 4 March 2019 https://doi.org/10.22452/MJES.vol56no1.4 ===== PAGE 11 (score 15, numeric-paren 0, table-words 2, var-words 9) ===== Malaysian Journal of Economic Studies Vol. 56 No. 1, 2019 73 Hedonic Analysis of the Impact of Flood Events on Residential Property Values in Malaysia houses are selected for this study. Meanwhile, distance variables such as distance to nearest bus stop, distance to nearest city centre, proximity to airport and proximity to school are obtained from Google Maps (Google, n.d.). 4.0 Results and Discussions A total of 1,486 respondents were interviewed in this study. However, only about 91.52 per cent or 1,360 respondents responded to the interview; the remaining 8.48 percent or 126 respondents declined to be interviewed due to time constraints. A majority of the respondents have busy work schedules and irregular working hours, which makes it difficult for them to accommodate our interview session. In particular, 700 respondents were property owners who had experienced flooding. The list of flood victims who still resides in the same areas that had been flooded was acquired from the Department of Social Welfare and from the local councils. The remaining 660 respondents who were not affected by flood in the area were also included in this study to avoid sample selection bias problem. Table 6 presents the descriptive statistics of the variables. It shows that at most, the study areas have been flooded three times within a five-year period (2008-2013). According to the Department of Irrigation and Drainage, most of the flood prone areas covered in this study experienced three flood events a year. For example, Larut Matang and Selama (one of our sample districts) were flooded five times in 2010; once in August 2010, three in September 2010, and another one in November 2010. The average cost of structural damage to the respondent’s house due to the flood event is RM966 including damage to the external and internal house structures. Based on the survey responses, the types of house structure damages recorded are paint cracks on Table 6. Descriptive statistics of variables used in the first stage and second stage of residential property value model Variable Mean Minimum Maximum RPV (in RM per unit property) 266,080.73 69,600.00 590,000.00 Flood: FREQUENCY (number of flood 3.13 1.00 5.00 occurrences in 5 years) DAMAGE (RM) 966.00 200.00 8000.00 Location: BUS (in kilometres) 2.36 0.70 10.60 CITY (in kilometres) 3.78 0.80 11.00 Structural: LOT (per square foot) 1,516.73 1,200.00 1,600.00 AGE (number of years) 15.00 8.00 20.00 BED (number of rooms) 3.13 1.00 7.00 BATH (number of rooms) 2.04 1.00 4.00 Source: Authors’ survey. ===== PAGE 12 (score 27, numeric-paren 0, table-words 5, var-words 17) ===== 74 Malaysian Journal of Economic Studies Vol. 56 No. 1, 2019 Nur Hafizah Ismail, Mohd Zaini Abd Karim and Bakti Hasan-Basri the wall, spoiled flooring, jammed or broken windows and doors, and some wreckage of the electrical system including switches and wiring. This study covers both low cost and high cost properties where the sale prices or residential property values (RPV) are between RM69,600 to RM590,000. The significant variations between the low cost and high cost property prices are due to the differences in the study areas which cover the urban and rural areas. The residential properties included in this study are located about 0.7 km and 0.8 km from the nearest bus station and city centre, respectively. Lot sizes of the sample residential properties can be considered moderate to large, with lot size averaging 1,516 sq. ft. The age of the properties surveyed range from 8–20 years. At most, the properties have seven bedrooms and six bathrooms. 4.1 First-stage Hedonic Regression The first objective of this study is to investigate the effect of flood frequency on RPV in urban and rural areas in Malaysia. For this purpose, we use the semi-log functional form (dependent), which can correct heteroscedasticity and multicollinearity problems (Basu and Thibodeau, 1998). Table 7 presents the result of the first-stage hedonic regression. Based on the results in Table 7, flood frequency in both urban and rural areas are negatively and statistically significant. The results explain that if flood frequency increases by one flood event per year, the residential property value significantly decreases by 18.5 percent and 13.6 percent in the urban and rural areas, respectively. The value of flood coefficient in the urban area is higher compared to the estimate in the rural area. This indicates that flood frequency causes the residential property value in the urban area to decline more than the property value in the rural area. Most of the urban residents are wealthier than the rural residents where the average income of urban residents is RM2,000 more than the average income of rural residents (DOSM, 2013). Therefore, urban residents can afford to pay a higher price for residential property in the urban area whereby most of the property has more valuable extras (NAPIC, 2014). Thus, floods that hit the urban area will have a greater effect on the urban property value compared to the rural area. Location is another important attribute that affects potential buyers’ housing choices. The urban-area CITY coefficient is positive and statistically significant but the coefficient of CITY2 is negatively significant implying that residential properties located near the city centre of urban areas have lower market values than those that are farther from the city centre. However, up to a certain distance, the farther the distance from the city centre, the lower the value of the residential properties. Buyers prefer to acquire residential properties farther away from the city but to a certain extent – especially away from industrial zones – because such city and industrial areas are more exposed to air pollution, which is hazardous to health (Schwarz & Marcus, 1990). Thus, this causes low demand for residential properties located in the city and its surrounding areas, thereby affecting their property market value. The structure of a house can be analysed in terms of lot size, house age and number of bathrooms and bedrooms. Coefficients for the lot size (LOT), number of bedroom (BED) and number of bathroom (BATH) for both urban and rural areas are ===== PAGE 13 (score 78, numeric-paren 41, table-words 10, var-words 12) ===== Malaysian Journal of Economic Studies Vol. 56 No. 1, 2019 75 Hedonic Analysis of the Impact of Flood Events on Residential Property Values in Malaysia positive and significant. An increase in the lot size usually adds to the value, because it gives the property owners room to expand such as adding more bedrooms and bathrooms. In fact, building additional rooms (bathroom or bedroom) will increase construction costs, which will consequently increase the value of a residential property. The coefficients of house age (AGE) for both urban and rural areas are negative and statistically significant. Generally, as a property gets older, it becomes less valuable because a lot of work will be needed for repairs and upgrading the condition as well as to maintain it. In fact, based on the assumption that older homes may have outdated technology as well as lacking a number of features available in newer homes, this would decrease the benefits provided by the property. Therefore, this will lower the value of the property. In the neighbourhood characteristics, all of the urban and rural areas DAIRPORT coefficients show insignificant results. Conversely, coefficient of proximity to school (DSCHOOL) shows a positive and significant result only in the rural area. This probable Table 7. Estimation results of the first-stage hedonic regression Coefficient Variable Urban area Rural area Coefficient Standard error Coefficient Standard error Constant 11.553*** (0.173) 11.336*** (0.216) Flood: FRE -0.185*** (0.070) -0.136* (0.081) Location: BUS -0.016 (0.011) 0.003 (0.006) BUS2 0.0001 (0.0001) -0.0001 (0.0003) CITY 0.041*** (0.012) -0.010 (0.008) CITY2 -0.001*** (0.0003) 0.0003 (0.0003) Structure: LOT 0.0001*** (0.00005) 0.0001*** (0.00004) AGE -0.012*** (0.003) -0.010*** (0.002) BED 0.077*** (0.034) 0.059** (0.045) BATH 0.233*** (0.035) 0.180*** (0.042) Neighbourhood: DAIRPORT 0.059 (0.074) -0.112 (0.090) DSCHOOL -0.139 (0.094) 0.256* (0.137) No. of observations 680 680 R2 53.90 47.00 Adjusted R2 52.40 45.30 F-Statistic 36.57 27.811 (0.000) (0.000) Note: *, **, and *** refer to figures that are significant at 10%, 5%, and 1% levels of significance, respectively. ===== PAGE 14 (score 8, numeric-paren 0, table-words 2, var-words 4) ===== 76 Malaysian Journal of Economic Studies Vol. 56 No. 1, 2019 Nur Hafizah Ismail, Mohd Zaini Abd Karim and Bakti Hasan-Basri reason may be due to the limited number of schools in the rural area while a number of them are located far from certain residential areas. For example, in Selangor, a rural area such as Kapar only has ten primary schools and two secondary schools, compared to the number of schools in urban areas, such as the city of Klang, which has a total of 75 primary schools and 36 secondary schools (Ministry of Education Malaysia, 2014a; 2014b). For these reasons, buyers in the rural area prefer to choose a residential property located near to a school. 4.2 Results of the Second-stage Regression The second objective of this paper is to determine the consumer’s willingness to pay for flood control measures to reduce the impact of flood risk. Implicit price measures the amount of money that individuals are willing to pay to keep their utility at original levels. The economic welfare that we used in this study involves a household’s comfort and satisfaction of their current living standards of being able to avoid the damages and inconveniences brought by flooding. Thus, in our study, implicit price refers to the estimates of economic welfare of households, as measured by their willingness to pay for flood mitigation measures to reduce the flood risk. The calculation of implicit price represents an empirically implementable measure of consumer’s willingness to pay for flood mitigation project to reduce the impact of flood risk. The implicit price can be calculated by using the formula as in Equation (4)3: (4) The total implicit price provided by respondents is calculated as RM4,461,878 for approaches to avoid losses in residential property value due to flood risk problems. The estimation of the individual amount can be calculated by dividing the total implicit price with the number of respondents. On average, the individual respondent’s implicit price for a flood management programme to reduce the flood risk problems in their residential area is about RM2,280. The total value of implicit price of RM4,461,878 or RM4.4 million calculated corresponds with the cost incurred by the government in the implementation of several flood mitigation projects such as river widening and deepening work, upgrading and modifying infrastructure and irrigation system and river maintenance work. For example, based on the report from the Perak Department of Drainage and Irrigation 2015/2016, the total cost of the flood mitigation project and the maintenance of several rivers in Perak such as Sungai Bidor and Sungai Lamin is about RM4.1 million. The close value between the flood control project’s actual cost and the estimated amount of households’ implicit price shows that this study can provide a good estimate Implicit Price RPV FREQUENCY = ∂ ∂ = coefficient of FREQUENCY FREQUENCY RPV × 3 For example, for observation 1 (or sample 1), suppose the residential property value (RPV) is RM100,000, coefficient of flood frequency (FRE) is 0.185 and variable flood frequency (FRE) is 3 times. The implicit price using the formula would be (0.185/3) x 100000 = 6166.67. Therefore, the implicit price is RM6,166.67. To compute the total implicit price, we need to add all the implicit prices for each observation. ===== PAGE 15 (score 29, numeric-paren 5, table-words 8, var-words 7) ===== Malaysian Journal of Economic Studies Vol. 56 No. 1, 2019 77 Hedonic Analysis of the Impact of Flood Events on Residential Property Values in Malaysia for flood management policy implementation, where the government can collect money from households for flood mitigation purposes rather than fully financing the flood project, since the respondents are willing to spend more money on flood mitigation projects in order to reduce the flood risk problems. The purpose of estimating the second stage hedonic regression model where the measure of willingness to pay is included as an dependent variable is to see the effect of socio-economic factors and risk on the willingness to pay to reduce flood risk. Table 8 presents the results of the second-stage hedonic regression. The positive and statistically significant coefficient of risk taking (DMOVE) shows that respondents who are risk averse, indicated by those who are planning to move to other flood free location, are willing to pay more money compared to those who are risk takers. A study done by Cameron and Shah (2015) found that natural disasters can impact real life behaviour through decreased risk taking behaviour. Commonly, after the post flood events, developers usually do not take on the full risk of responsibilities of the property in areas that are prone to flooding. Therefore, for the flood risk adverse group, they are willing to pay higher for flood control measures to reduce the impact of flood risk, so as not having to move from the area. Table 8. Estimates of the second-stage hedonic regression Variable Coefficient Std. deviation Constant 8.414*** 0.229 Flood DAMAGE 0.00046 0.0039 DMOVE 0.357*** 0.115 (1= planning to move; 0 = not moving) Socio-economic profile DINCOME 0.132* 0.096 (1= above RM3,050; 0 = below RM3,051) DEDUCATION 0.347** 0.137 (1= university; 0 = otherwise) AGE -0.010** 0.004 DFINCON -0.065 0.043 No. of observations 700 R2 47.00 Adjusted R2 39.00 F-Statistic 15.67 Note: *, **, and *** refer to values that are significant at 10%, 5%, and 1% significance levels, respectively. ===== PAGE 16 (score 22, numeric-paren 0, table-words 5, var-words 12) ===== 78 Malaysian Journal of Economic Studies Vol. 56 No. 1, 2019 Nur Hafizah Ismail, Mohd Zaini Abd Karim and Bakti Hasan-Basri However, the value for the coefficient of DAMAGE is statistically insignificant. This shows that the cost of structural damage to the house due to a flood event does not affect the respondent’s willingness to pay for flood control measures. This may be due to the fact that most of the respondents in this study had experienced flooding more than once. As such, they have made preparations such as installing temporary barriers and upgrading permanent defences and resilient equipment on the house structure to reduce damage and loss from floods. In fact, based on the survey (as presented in the descriptive statistics), the average house structural damage cost incurred by the 631 respondents who experienced flood events more than once is small (even under RM1000). Therefore, the cost of structural damage to the house due to a flood event does not significantly affect the decision of respondent’s willingness to pay for flood control measures to reduce flood risk. Other socio-economic factors such as income, age and education level have in- fluenced the level of willingness to pay (Devkota et al., 2014; Maraseni et al., 2008). The coefficient of household income (DINCOME) is positive and statistically significant. Generally, when household income increases, households’ spending ability will improve, thereby improving their economic welfare. This means that they would be willing to pay to maintain the same level of economic welfare. Based on the Malaysian Household Income and Basic Amenities Survey Spending Report 2013, there is an increase of 6.13 percent in household income over two years between 2012 and 2013. Therefore, with the increase in income and spending ability, residents willingness to pay for flood management programmes also increases. A positive and statistically significant coefficient of education level (DEDUCATION) indicates that households that possess a university qualification tend to have a higher willingness to pay for flood reduction programmes compared to those households with- out a university education. This may be because university graduates are more exposed to environmental issues (i.e., flood disasters), which would make them more likely to be more concerned and knowledgeable about flood and its management programmes. Thus, they would be more capable of assessing flood risks and the potential damage to properties as compared to those with a lower educational attainment. This is also consistent with the findings of an income report where according to the Public Services Commission of Malaysia, the starting salary of a university graduate is on average RM2,500 higher than those with a lower educational attainment. Therefore, this would make the university graduate respondents willing to spend more money for flood control measures due to their higher financial capacity. Based on the result, the coefficient of age (AGE) is negative and statistically significant. As age increases, some of the older households who are phasing into retirement may deal with lower pay and also some are likely to retire early due to poor health. As a result of the uncertainty and lower income received by the older household, this may affect their willingness to spend for flood control measures to reduce flood risk. Hence, increase in age will reduce the household willingness to pay for flood management programmes. On the other hand, the coefficient of financial constraint (DFINCON) is statistically insignificant. This shows that the effect of respondent’s income on willingness to pay for flood control measures is not influenced by the respondent’s decision to move or not from the flood prone area (DMOVE).