Low-elevation coastal zones (LECZs) are home to more than 10% of the world’s population, and this share is growing (Neumann et al. 2015). But the attraction of LECZs poses problems, due to their susceptibility to floods, including from tropical storms, which affect many millions of people around the world every year (Brakenridge 2021). The problem of flooding is acutely felt in the US, where in the present conditions, expected annual losses from tropical storms alone are estimated at $57 billion, of which a third is borne by taxpayers (US Congressional Budget Office 2019). To make matters worse, global sea levels are rising (Pörtner et al. 2019), with varying consequences for different coastal areas: the US Atlantic and Gulf coasts have experienced some of the fastest local rates of sea level rise (SLR) in the world during the 20th century. This trend is expected to continue, increasing the frequency of flooding – in some cases by an order of magnitude (Dahl et al. 2017).
Against this backdrop, it is important to understand where housing construction in coastal areas is taking place. In recent research (Lin et al. 2021), we assemble a new dataset on the location of housing and flood risk across thousands of kilometres of coast, spanning major urban centres, small towns, and rural areas for the US Atlantic and Gulf coasts, to investigate this. The data, which cover two decades, are at a highly disaggregated spatial scale. They include information on housing from the census and land cover from satellite imagery, as well as measures of SLR proneness, flood damages, and regulatory restrictions. These data allow us to explore construction in areas where flood risks for residents and taxpayers are both high and rising, due to climate change. We use these data to document how the existing housing stock and new construction vary by distance to the coast. The result is a novel and detailed picture of housing in LECZs, and its relationship to the vulnerability of different locations to flooding and SLR.
We also develop a model,that provides a parsimonious explanation for our findings. The model answers questions such as: Why does housing concentrate near, but not right at, the coast? Why are coastal cities asymmetric? Why is new housing in LECZs increasingly built on flood-prone areas, which were previously avoided? And why does this happen especially on the urban fringes? Finally, we extend our model and use it to study how SLR may reshape cities, and consider implications for rising costs of flooding and taxpayer subsidies, the economic decline of some neighbourhoods, and lengthening commutes.
We begin by documenting nine stylised facts, which we group into three broad take-away findings:
The Shape of coastal locations in 1990
- First, housing unit density peaks near, but not right at the coast, and it declines more steeply on the coastal side.
- Second, coastal places are asymmetric – their central business district (CBD) is closer to the coast side edge.
- And third, this asymmetry increases in city size.
Figure 1 Population concentrates near – but not right at – the coast
Notes: This figure shows coefficients (bold line) and 95% confidence intervals from a regression of the logarithm of housing units per square kilometer on distance bins from the coast using more than 400,000 census blocks within 10 km of the US Atlantic and Gulf coasts. The housing density distribution peaks near, but not right at the coast. It declines more sharply on the coast side.
Figure 2 Places near the coast are asymmetric
Notes: This figure illustrates the asymmetry of coastal places by showing census-designated places in the vicinity of Boston, with their CBDs in circles. Places close to the coast are asymmetric, with CBDs close to the coast side, while places further inland are more symmetric around their CBDs.
What explains the shape of coastal locations?
- Fourth, census blocks that are prone to SLR are much more sparsely built; but conditional on SLR-proneness, blocks closer to the coast are more densely built.
- Fifth, SLR-proneness rises steeply near the coast.
- Sixth, damage from flooding also rises steeply as we approach the coast.
Together, these stylised facts suggest a tension between the amenity of coastal proximity, and the disamenity of flood-proneness, which increases steeply near the coast.
Figure 3 Damages from flooding decline rapidly with distance from the coast
Notes: This figure shows the logarithm of National Flood Insurance Plan (NFIP) claims over several decades, normalized by the number of housing units, by distance bins to the coast. The locations closest to the coast have losses per housing units that are 2-3 log points (or roughly 7-20 times) larger than locations further inland.
Patterns of growth from 1990-2010
- Seventh, net new construction from 1990-2010 was more than twice as prevalent in SLR-prone locations as in the 1990 stock of housing.
- Eighth, SLR-prone areas were more likely to be developed in dense census tracts.
- Finally, in the densest census tracts, new construction focused on medium-risk SLR-prone areas, while avoiding the riskiest ones.
Figure 4 Miami Beach and Miami, FL
Notes: This figure shows an area in Miami and Miami Beach, Florida. The highlighted census blocks have a positive share of land that is prone to a one foot rise in sea level, were in dense census tracts in 1990, and exhibited positive (yellow) or positive and large (orange) growth in net housing units from 1990-2010. We find similar patterns in other cities, such as New York, Boston, and Tampa.
We show that our nine stylised facts are robust to excluding census blocks, which were mostly shielded from private residential construction, because they are either protected areas, military bases, or parks. We also find evidence consistent with our stylised facts when we use data on built area, based on satellite imagery, which cover all construction rather than just housing data.
To account for the nine stylised facts, we develop a model of a monocentric coastal city. In the model, coastal areas are characterised by both an amenity, which declines linearly in the distance to the coast, and a disamenity (flood-proneness), which declines convexly in distance to the coast. The city founder chooses a location that trades off these two factors – close to the coast, but not right at it. This location becomes the city's focal point – the CBD. Residents then choose where to live, and they prefer locations close to the CBD, both because of their high net amenity value and because of the shorter commute. Housing density peaks around the CBD, but declines more steeply on the coast-side, because of the convex flood-proneness. The city expands over time into previously empty areas on both sides. On the coast side, this expansion involves building on increasingly flood-prone land.
Taken together, our nine stylised facts and our model of a coastal city, which helps us to account for what we observe in the data, suggest that coastal cities face ‘soft’, as opposed to ‘hard’ (Saiz 2010), barriers related to flood risk. Soft barriers are locations that are not used for housing development in most circumstances, but are nevertheless built on as cities expand. Construction on soft barriers – another example of which is construction in areas prone to wildfires – may involve risks not only to residents but also externalities (e.g. for taxpayers or the environment), which may necessitate policy intervention.
We also extend our model in several ways, including to allow for sea level rise and government subsidies to flood-prone areas. We then simulate our model to explore challenges that low-elevation coastal cities may face in the coming decades. These simulations point to four potential concerns for low-elevation coastal cities. First, the problem of housing in flood-prone locations looks set to worsen, either because cities expand towards the coast, or because of SLR, or because both happen simultaneously. This development threatens to increase flooding costs for both residents and taxpayers. Second, even if LECZ cities grow on aggregate, some neighbourhoods within them may experience economic decline, as increased flood risk causes demand for housing to decline. This problem is exacerbated in the case of economically stagnant cities. Third, SLR further distorts the shape of LECZ cities, leading to increasingly “misshapen” coastal cities (Harari 2020), and significantly lengthening the time costs of commuting to work. Finally, these cities face a potential crisis if their CBD comes under threat of being permanently submerged.
Figure 5 Simulation results
Notes: Simulation of an initially small city (in yellow), which becomes more asymmetric around its CBD (black vertical line) as it expands towards the coast. Expansion leads to increasing exposure to flood risk on the coast-side and lengthening commutes on the inland side. The expansion towards the coast takes place despite rising sea levels (in blue, expanding gradually inland). This expansion worsens the city’s asymmetry.
Notes: Here we consider a dynamic version of the model, where expansion into previously unbuilt land is costly and irreversible. This figure shows an economically stagnant city. As sea levels rise, some parts of the city are permanently submerged, while areas that are not yet submerged become less attractive and decline economically (in red).
Governments could enact various policies to mitigate the problems we highlight, especially the rising exposure of taxpayers. First, governments could consider taxing new developments in flood-prone areas, if there are viable alternative uses to the land which are not taxed. The limiting case is an outright ban on extensive margin developments, although enacting and enforcing such a ban might be difficult.
Second, governments could offer the subsidy only to existing housing. One such policy is the UK government’s Flood Re, which provides subsidised flood insurance only to ‘grandfathered’ housing, built before 2009. Governments could also attach further conditions to their subsidy. These conditions could include stricter building standards, such as construction on stilts imposed by the US Federal government when compensating the victims of Hurricane Sandy. Or governments could restrict the number of times a given property is bailed out, or offer other incentives to move instead of rebuilding, as Canada has recently done.
With SLR proceeding at pace, the costs to taxpayers of fixing neighborhoods or even cities may at some point become prohibitive. An example of how far things have deteriorated in another part of the world can be seen in Indonesia, whose government is investing heavily in moving its capital from flood-prone Jakarta. Ultimately, of course, slowing down climate change and SLR could also reduce the costs, especially those associated with large-scale urban moves. This remains a central policy challenge.
Our paper offers a path for researchers and policymakers to consider the implications of a range of interventions in low-elevation coastal cities, not just on the US Atlantic and Gulf coasts, but for many coastal areas around the world, including in developing countries where LECZs are experiencing massive population increases. All of this in an era when climate change poses increasingly important challenges.
Brakenridge, R (2021), "Global active archive of large flood events", Dartmouth Flood Observatory, Boulder, CO: University of Colorado.
Dahl, K A, M F Fitzpatrick, and E Spanger-Siegfried (2017), "Sea level rise drives increased tidal flooding frequency at tide gauges along the U.S. East and Gulf Coasts: Projections for 2030 and 2045", PloS ONE 12(2).
Lin, Y, T McDermott, and G Michaels (2021), “Cities and the sea level”, CEPR Discussion Paper 16004.
Neumann, B, A T Vafeidis, J Zimmermann, and R J Nicholls (2015), "Future coastal population growth and exposure to sea-level rise and coastal flooding - A global assessment", PloS ONE 10(3).
Pörtner, H O, D C Roberts, V Masson-Delmotte et al. (2019), "IPCC special report on the ocean and cryosphere in a changing climate", United National Inter-Governmental Panel on Climate Change.
Saiz, A (2010), “The geographic determinants of housing supply,” The Quarterly Journal of Economics 125(3): 1253-1296.