1 Introduction

Over 8 million homes globally are projected to be destroyed within the next 20 years if extreme climate events continue to rise (Seabrook 2021). Concerns about how to adapt to the impacts of extreme weather events are rapidly increasing especially for the housing market. Arguably, to be somewhat successful in adaptation requires quantifying climate-induced risks and losses to strengthen adaptation policies. Jamaica, one of the most vulnerable countries in the Caribbean, is at risk of losing its homes. The country has been plagued by innumerable disasters, which has drawn the attention of researchers in not only quantifying the impact of these events but also predicting future losses and highlighting the need to implement adaptation techniques (Smith and Mandal 2014; Burgess et al. 2018; Spencer and Urquhart 2021; Collalti and Strobl 2022).

Like many other countries within the region, the country’s policymakers are concerned with adaptation. However, one aspect that has received no attention is the housing market. Given post-disaster assessments that survey quantitative losses, it is clear that this market is unable to withstand the impacts of extreme events. For example, a compilation report calculated USD 351 million in damage done to residential buildings for the period 2001 to 2012 from intense rainfall and hurricane events (Smith and Mandal 2014). Further, the heavy rains in 2021 caused by tropical storm Elsa resulted in flooding of many homes and approximately USD 6 million in damage (Gibbs 2021; Linton 2021). Such damages may result in perceptional changes of the risk of locations for housing and thereby affect prices and possibly investment in further housing construction (Fang et al. 2021). There is also some indication that extreme events may increase with climate change in Jamaica (Knutson et al. 2020; Vosper et al. 2020). There is thus a need for research to quantify the impact of extreme climate events on the real estate market to accurately inform policy decisions on preparation and adaptation.

As indicated above, there is no known research quantifying the impact of extreme climate events specifically on the Jamaican real estate market. Nevertheless, Apergis et al.’s (2020) cross-country study included Jamaica as part of a panel of 117 countries and estimated a lowering of house prices as a result of natural hazards and disasters. A similar directional impact is estimated for specific countries and extreme events throughout the literature. In some recent studies, Rajapaksa et al. (2016), Zhang (2016), and Atreya and Ferreira (2015) found that flooding has a dampening effect on property values in Australia, and the North Dakota-Minnesota metropolitan area and Albany, Georgia in the United States, respectively. Further, van Veelen (2020) reported discounted sale prices on residential properties in Orange County, Florida in the United States following the 2004 strike of hurricane Charley. Such discounted sales would lead to significant reductions in income for property owners as shown by McAlpine and Porter (2018) who estimated USD 465 million in losses for the real estate market in MiamiDade, Florida. On the contrary, hurricanes and winter weather have been shown to increase house values, albeit in the short term, for coastal cities in the United States and Colorado respectively (Murphy and Strobl 2009; Gourley 2021). Even though a majority of the literature leans towards a property value reducing effect of extreme climate events, there is still a possibility that property values may increase as Murphy and Strobl (2009) and Gourley (2021) demonstrated.

The real estate and disaster literature, although seemingly a growing one and has taken important steps in demonstrating the potential impacts of extreme weather events, is arguably somewhat deficient in some respects. First, Latin America and the Caribbean (LAC) is an under-studied region even though it is classified as the world’s second most disaster-prone region (United Nations Office for the Coordination of Humanitarian Affairs 2020). For example, in the last 20 years, the region experienced over 1200 disasters, of which over 45% and 27% were floods and hurricanes respectively. In terms of specific countries, Jamaica, which has not been studied, is cited as one of the top two countries that have attained a high-ranking status for being physically exposed to tropical cyclones. Second, the large majority of studies is focused on flooding. Although this is the most common type of natural hazard, tropical cyclones generate greater economic damages globally and deserves due attention (World Meteorological Organization 2021). Third, Apergis et al.’s (2020) study, although including 18 LAC countries of which Jamaica is listed, aggregated geophysical, meteorological, hydrological, climatological, and biological disasters from the International Disaster Database (EM-DAT). This aggregated measure fails to disentangle the impact of the main disasters affecting countries in the region. Further, EM-DAT is not the most reliable database for disaster data given the different sources from which data were collected (George et al. 2021).

Jamaica is particularly useful in addressing the shortcomings of the current literature on the LAC region for several reasons. First, it is frequently struck by flooding events induced by excess rainfall, the intensity of which is predicted to increase in the future (Collalti and Strobl 2022). Data available from the Jamaican Office of Disaster and Preparedness Management show that, between 2001 and 2018, excess rainfall resulted in more than 1000 locations across the island being affected by almost 100 flash-flood events lasting between one to 14 days (Collalti et al. 2023). Second, it also experiences a lot of hurricanes. Data from the Meteorological Service of Jamaica reveal that the island experienced the effects of at least 60 hurricanes since 1900. Third, one can isolate the effects of different types of disasters on real estate values.

The main aim of this study therefore was to quantify the impact on real estate in Jamaica arising from extreme rainfall and hurricanes. To this end, an exhaustive geolocalized database on mortgage and land and apartment sale values was combined with localized rainfall and hurricane track data. The quantification produced some important findings. The average hurricane reduces apartment sale prices by at least 50% but do not affect the value of mortgages taken and residential land sale prices. Further, while extreme rainfall increases the value of apartments by 44%, it reduces the value of mortgages by roughly 15% but had no impact on residential land sale prices.

The results of this study are particularly useful for targeting economic inequality given that property ownership represents a major portion of households’ wealth holdings and is a symbol of financial security (Goodman and Mayer 2018). In addition, as the results show, extreme weather events have the potential to reduce property values, which may encourage households to become renters rather than owners, where in the former role they would not be responsible to bear the financial burden for repairs.

The rest of this article is organized as follows. Section 3 describes the data and method of estimation. The results and their discussion are provided in Sect. 4. Section 5 concludes.

2 Background

Jamaica has a long history of dealing with extreme climate events including hurricanes and extreme rainfall. These events generally result in minor to severe damages to housing with assessments reporting high replacement costs despite the majority of structures being built with strong building materials.

Jamaica’s first documented experience with hurricanes dated back to as early as 1,559 when a hurricane strike caused considerable damage to the country’s infrastructure (National Library of Jamaica 2012). Since then, the housing sector has suffered immensely. In earlier years, sources indicate that 9,000 homes were lost from hurricane Charlie (1951), 40,000 from the Gilda floods (1973), 5,000 from hurricane Allen (1980), and 500,000 from hurricane Gilbert (1988) (Norton 1952; Reyna 1988). The destruction caused by the hurricanes varied due to their strength and duration and whether they made landfall. In 2004, hurricane Ivan, a category 5 storm, reportedly damaged over 700,000 dwellings (Planning and Institute of Jamaica 2004). In 2012, hurricane Sandy, a category 1 storm, damaged over 16,000 homes (Planning Institute of Jamaica 2013).

The governmental reports do not present regular assessments of extreme rainfall thus data on their damage costs are limited. Nevertheless, the 2010 flooding assessment reported over 2,169 homes being impacted by heavy rains (Planning Institute of Jamaica 2010). Evidence also points to damages to homes from more recent floods occurring between 2016 and 2021 (Davies 2016a, Davies 2016b, Davies 2016c, Davies 2017a, Davies 2017b). On a cautionary note, the damage costs presented are based on estimates derived from post-disaster surveys involving residents from affected areas. It is unclear whether all affected properties were included and so the above-mentioned estimates may be an understatement of the true costs.

3 Data and Estimation Strategy

This section presents the data used and the summary statistics for all variables. In addition, it discusses the estimation strategy employed for the analysis.

3.1 Data

The source of the property sales data is the Jamaican National Land Agency.Footnote 1 This database is exhaustive for 2003 to 2018. These data are annual deflated individual residential land and apartment sale prices and mortgage values. They are collected across the 14 parishes in Jamaica, providing at least 172,000, 15,000, and 178,000 observations for residential land sales, apartment sales, and mortgages respectively.

For rainfall data, the Global Precipitation Measurement (GPM) satellite mission database,Footnote 2 which provides half-hourly observations at the 0.1° resolution, was used, with high resolution rainfall averages measured in mm. Excess rainfall was identified when the daily rainfall exceeded the 90th percentile of a grid’s non-zero values over the period of study.

For hurricanes, data from the National Hurricane Center’s databaseFootnote 3 were used with Boose et al.’s (2004) form of Holland’s (1980) wind field model to construct proxies of localized strong wind damage. Importantly, to convert wind speed into potential damage, the cubic power of wind speed was used since it has been established that the relationship between property damage and a hurricane varies with the cubic power (Emanuel 2005; Strobl 2012; Ishizawa and Miranda 2019; Spencer and Strobl 2019). This approach to calculating hurricane wind damages is now in the climate change literature and is regarded as advantageous since it accounts for the physical characteristics of a storm that determines the spatially heterogeneous nature of wind speeds experienced locally, for example, the position of residential properties relative to the storm, the maximum wind speed and the movement of a storm, and whether landfall was made. Moreover, the approach is more superior than using storm incidence dummies or ex post estimates of destruction (Henry et al. 2020). Consequently, a hurricane wind damage index (H) that takes into account how storm features impact wind speed locally is defined as:

$${H}_{ijt}=\sum_{q=1}^{Q}\sum_{l=1}^{L}{W}_{iqlst}^{3}\dots {W}_{iqlst}\ge {W}_{ij}^{*}$$

where i = 1,..., I properties located in parish j = 1,..., J experience hurricanes q = 1,..., Q with lifetime of l = 1,...,L. Note that W is the local wind speed and W* is the thresholdˆ above which the wind has destructive power. The threshold choice is 119 km/hour, which corresponds to the definition of a category 1 hurricane as defined by the Saffir-Simpson Scale and has been noted by Henry et al. (2020) as the speed at which residential properties in Jamaica start to experience destruction.

3.2 Summary Statistics

Table 1 displays the descriptive statistics for the data used in the analysis. The average sale prices for land and apartments are USD 57,988 and USD 64,465 respectively while mortgages taken average USD 48,145 over the study period. Given that the rainfall and hurricanes are location specific, these values are different for all three types of real estate. Thus, the average extreme rainfall values are 0.0203 (residential land sale prices), 0.0047 (apartment sale prices), and 0.0177 (mortgages). Finally, the average non-zero value of the hurricane indexes, that is, when they are destructive, is 0.0072, 0.0069 and 0.0072 in order of their respective listing in Table 1.Footnote 4

Table 1 Summary statistics

3.3 Estimation Strategy

To estimate how climate events may affect property sales across years and parishes, a linear fixed effects model was used. This model is standard in the hedonic pricing literature (Beltrán et al. 2019; Mutlu et al. 2023; Saptutyningsih and Dewanti 2023) that examines the impact of climate events on economic outcomes. The model is as follows.

$$\rm{ln}\left({Y}_{ijt}\right)= \propto + \sum_{l=0}^{L}{\delta }_{{R}_{t-1}}{R}_{ijt-1}+\sum_{l=0}^{L}{\beta }_{{H}_{t-1}}{H}_{ijt-1}+{\pi }_{t}+{\mu }_{i}+{\varepsilon }_{ijt}$$

where i indicates property value, j parish, and t the year. Y is land sale prices, apartment sale prices, or mortgages taken. In line with the literature, Y is assumed to be the natural log of land, apartment, and mortgage values in each parish overtime. R and H constitute measures of extreme precipitation and wind exposure respectively and are described in Sect. 3.1. One should note that any persistent effects of these extreme climatic events are explored using their lagged values. In addition, π and µ capture year and parish fixed effects and \(\epsilon\) is the error term. After controlling for fixed effects, one can consider H and R to be random realizations of their assumed time invariant distributions and thus exogenous.

4 Results and Discussion

The study first investigated the individual impact of extreme rainfall and hurricanes on residential land sale prices. As shown in Table 2, neither excess rainfall nor hurricanes affect residential land sale prices significantly. The estimated rainfall outcome is in congruence with a strand of the flood risk and property literature that also found no effect on property (see for instance, Bin and Landry 2013; Cupal 2015, Murfin and Spiegel 2020). Specifically, Sawada et al. (2018) found no significant impact on residential land prices despite notable substantial damages after flood shocks in Thailand. According to Sawada et al. (2018), non-significance of estimates possibly relates to illiquidity or other interferences that take place in the real estate market. On the other hand, some studies indicate significant price reductions after flooding especially for properties with higher flood risks (Zhai et al. 2003; Ismail et al. 2016; Dudzińska et al. 2020; Wei and Zhao 2022). In the case of residential land in this study, the data do not indicate property attributes such as proximity to streams or rivers, the condition of drainage systems close by, the slope, or the extent of deforestation and soil drainability that are factors that can induce a higher risk of flooding. In this regard, one can only assume that characteristics such as these could influence the estimated outcomes. As it relates to hurricanes, the literature does not present any studies on residential lands. Although this relationship remains unexplored, one can assume that this non-existent hurricane impact is reasonable since the construction of the hurricane variable only captures wind destruction and with no physical structure to destroy, the land values would remain unaffected.

Table 2 The impact of climate extremes on residential land sale prices

In Table 3, the estimates show that both types of climate extremes impact the sale prices of apartments where their directional effects act independently as is shown beyond models 1 and 2. Interestingly, we observe that extreme rainfall commands an increase in apartment prices by roughly 45%. However, this effect is only short term and disappears in the year following the shock. It is not surprising that this directional impact is generally at odds with the majority of the literature. Nevertheless, one finds indirect support from Murphy and Strobl (2009) and Gourley (2021) where contemporaneous increases in property values are estimated albeit for two alternate climate extremes, cold weather and hurricanes. However, in the case of extreme rainfall, one can adopt Murphy and Strobl’s (2009) rationale where an excessive amount of rain reduces the supply of livable properties more than they reduce the size of the population in specific localities, which in turn drives up their prices. One can also assume another possibility based on Gourley (2021), that is, potential buyers see fewer apartments in adverse weather so that they lack enough information to negotiate when they are purchasing and as a result, they pay higher prices because sellers have a stronger selling position. Along with Murphy and Strobl (2009) and Gourley (2021), the positive impact of weather extremes on property values estimated in this study stands in contrast to the extensive literature that presents discounted values resulting from flood incidences (see for instance, Zhang 2016; Beltrán et al. 2019; Apergis et al. 2020; Miller and Pinter 2022).

Table 3 The impact of climate extremes on apartment sale prices

Table 3 also shows that in contrast to the contemporaneous impact of extreme rainfall, the average hurricane reduces apartment sale prices by over 50%. This negative impact is quite common in the literature for house prices (for instance, Murphy and Strobl 2009; Ortega and Taspınar 2018). Although could not be proven in this study, possible reasons for such a large decline include the significant damage to structures that attract considerable resources amounting to more than the value of properties, people are pushed to migrate to avoid future disasters and need to quickly liquidate properties, or there is an increase in the risk perception of living in particular areas that home buyers or real estate agents use to negotiate prices downwards (Ortega and Taspınar 2018; Spencer and Urquhart 2018; Cohen et al. 2021). Further, research such as Below et al. (2017) and Komarek and Filer (2020) highlight that homes located in high-risk areas are likely to stay on the market a longer time period, which possibly indicates that in the case of the observed significant reduction in the sale prices of apartments in this study, sellers wanted to avoid the waiting period to sell. As it relates to the housing market literature, the strand that considers hurricanes is quite small; nevertheless, it supports the value reducing effect estimated in this study where properties suffer a fall in values (see for instance, van Veelen 2020, hurricane Charley in Florida, 1.9% to 2.4% discount; Below et al. 2017, multiple hurricanes in North Carolina, 3.8% discount). It is worth pointing out that the smaller discount estimates noted in these studies could possibly result from the use of storm and damage indicator variables whereas this study used a hurricane index. The results indicate the presence of a persistent effect two years after a hurricane strike, that is, a smaller reduction in apartment sale prices, which is around 40%. Regarding this smaller value reducing effect, it is possible that repairs would have taken place within the year of the storm strike so that apartment values do not drastically fall as they do initially.

Table 4 shows the results from the model estimating weather extremes’ impact on the value of mortgages taken. As with apartment sale prices, models 3 to 5 demonstrate that both extremes act independently in affecting mortgages. From the results, one generally observes a decline in the value of mortgages taken by approximately 24% from the impact of extreme rainfall. The results of models 4 and 5 also show that the addition of lags is insignificant thus indicating that the impact of extreme rainfall does not persist after the current period. The absence of a persistent effect could possibly result from mortgage loan agencies competing for an increase in customers and revenue.

Table 4 The impact of climate extremes on mortgages

This reducing effect is expected since the more at-risk locations would be denied mortgage loans; however, as pointed out by Keys and Mulder (2020), although in the case of sea level rise, there need not be a denial in mortgage loans but rather climate risk-induced pessimism on the part of potential buyers results in less mortgages being taken. Additionally, as the results show, mortgages taken do not continue to decline, suggesting that either mortgage institutions no longer have climate-related risk reservations or buyers are not pessimistic about the impacts of extreme weather. These explanations support others coming out of the mortgage finance-climate change literature where Ouazad and Kahn (2022), for example, provides evidence that mortgage institutions are more keen to approve securitized loans that transfers the climate risk to borrowers. Table 4 also shows that there is no hurricane impact. As literature directly investigating the impact of natural hazards and disasters on mortgages taken is non-existent, the results of this study will add a remarkable texture to the strand of climate change-real estate literature and lend support to the above-mentioned studies on climate-induced mortgage risk.

The mechanisms operating behind the negative and positive estimated outcomes in the general literature (for example, Atreya and Ferreira 2015; Gourley 2021), although unclear, can be attributed to various factors including changes in how buyers feel especially in bad weather conditions such as a cold winter where despite the availability of open houses, prospective homeowners postpone searching. Such postponement could increase prices since the buyers have less information leaving sellers with more negotiation power (Gourley 2021). However, depending on the type of weather, some buyers may not be deterred and as such continue house hunting and could as a result contribute to lowering prices (Gourley 2021). One could also consider that damages to nearby properties or facilities can also devalue the price of properties (Stull 1975). Further, decline in property values also could be driven by high reconstruction and psychological costs associated with extreme weather events (Atreya and Ferreira 2015). In the case of Jamaica, one can argue that the increase of apartment prices resulting from extreme rainfall can be attributed to buyers postponing house hunting as a result of bad weather so that upon resumption they have less bargaining power than sellers. Further, apart from the likelihood of a decline in buyers’ mood (Gourley 2021), they can also be affected by flooded roads and landslides. Another driving factor, albeit for reduced property values, could be the condition of roads, that is, ones that can become easily flooded or susceptible to damages or low-lying bridges that are in close proximity and are the only option for travel. These in effect are neighborhood assets that when affected by the extremes can reduce property values (Stull 1975). Further, both implicit (mental health) and explicit (reconstruction) costs could drive down prices if homeowners are unable to bear these costs, especially with no insurance, and decide to sell and migrate, internally or externally.

Table 5 provides robustness checks for the climate variables by considering a higher threshold for damaging hurricanes and higher threshold for extreme rainfall. Specifically, a threshold choice of 178 km/hour and daily rainfall exceeding the 95th percentile of a grid’s non-zero values are now used to re-estimate model 5 in Tables 2, 3, and 4. In general, the directional impact remains the same but the magnitude changes where for apartments, sale prices are now increased by almost 80% and reduced by roughly 86% for extreme rainfall and hurricanes respectively. Further, the model now estimates a smaller reduction (10%) in the value of mortgages taken. These differences in magnitude compared to the original model suggest that these alternative measures may not be accurately capturing the destruction of property.

Table 5 Robustness checks

5 Conclusion

In this study, the impact of extreme weather events on the Jamaican real estate market, as captured by the values of residential land sales, apartment sales, and mortgages taken was examined. The study found that extreme rainfall reduces the value of mortgages but increases apartment values. It also found that hurricanes reduce the values of apartments.

These results draw attention to one implication—wealth risk is involved in property ownership. In this regard, one should note that property ownership is a significant source of wealth and is expected to increase in value overtime. However, given that extreme weather events have been widely demonstrated to negatively impact real estate, the value of property ownership, albeit in the short term, can be severely eroded thus underscoring the need to pin down other sources of wealth or have in place financial sources of recovery such as insurance and personal savings, where the latter has been identified by Henry et al. (2020) as a coping mechanism for Jamaicans. However, mortgages from certain institutions require mandatory insurance. Thus, while under a mortgage, the destruction of properties will be covered by insurance payouts.

In terms of study limitations, the length of time rainfall events last would have allowed a better assessment of the risk of excess rain affecting property values. Unfortunately, no specific damage estimates are available for homes destroyed by extreme rainfall since available costs related to hurricanes and flooding are lumped together. In addition, the available property data do not contain the attributes of properties considered in standard hedonic models, which can affect the magnitude of the estimates that the literature has demonstrated. Having these would have added a nice texture to the article.