Report's Home Page Data and Supplemental Information

At a Glance

The federal government makes federally backed flood insurance available to people in areas that participate in the National Flood Insurance Program, or NFIP, by employing floodplain management practices that meet the program’s minimum requirements. Currently, 3.1 million properties (parcels of land used for any purpose) are covered by policies offered by the NFIP; that program provides almost all the nation’s flood insurance policies.

In this report, the Congressional Budget Office examines how the share of properties at risk of flooding that are covered by NFIP policies varies across communities with different economic and demographic characteristics. The agency considers properties to be at risk if they have at least a 1 percent annual probability of experiencing a flood of a depth of roughly 1 foot or more, which is equivalent to having about a one-in-four chance of experiencing at least one such flood over a period of 30 years; currently, about 9 percent of properties face such a risk.

CBO found that most at-risk properties did not have flood insurance through the NFIP as of May 1, 2023. Of at-risk properties with NFIP policies, about 90 percent had a discounted premium, and some of those discounts were time limited, or temporary. The agency also found the following:

Household Income. Properties at risk in communities in which median household income was lower were less likely to be covered by NFIP policies. Those policies were also less likely to have premium discounts.
Primary or Secondary Residences. At-risk properties in communities in which most dwellings were secondary residences were more likely to be covered by NFIP policies, and to be covered by policies with premium discounts, than those in communities in which most dwellings were primary residences.
Coastal or Inland Location. Properties at risk in coastal communities were more likely to have NFIP coverage and more likely to be covered by policies with premium discounts than were those in inland communities.
Other Characteristics. The likelihood that a property was covered by an NFIP policy also varied when communities were defined by the race or ethnicity of householders (generally, the person in whose name a residence was rented or owned), whether households included a senior or a child, and whether most residences were occupied by renters or homeowners. The variation among communities grouped by those characteristics was smaller than the other variations discussed above.

Notes About This Report

All dollar amounts for median household income are in 2020 dollars. To remove the effects of inflation, dollar amounts for years before 2020 are adjusted using the retroactive series of the consumer price index for urban consumers (CPI-U-RS). (That approach is consistent with the Census Bureau’s treatment of multiyear estimates, which are adjusted using the CPI-U-RS.) For more details, see Census Bureau, Understanding and Using American Community Survey Data: What All Data Users Need to Know (September 2020), p. 73, https://tinyurl.com/yw2cps7s.

In this report, the United States refers to all 50 U.S. states, the District of Columbia, and Puerto Rico.

In figures that report the values for quintiles, communities with the cutoff value are assigned to the higher quintile.

Some analyses in this report divide households into groups according to the householder’s race or ethnicity. The Congressional Budget Office groups householders into three racial groups (Black, White, and some other race or multiracial) and two ethnic groups (Hispanic or Latino and neither Hispanic nor Latino, simplified in this report to Hispanic and not Hispanic) on the basis of their self- identification as recorded in the American Community Survey’s five-year estimates. The racial groups specified in those estimates are American Indian and Alaska Native, Asian, Black or African American, Native Hawaiian and Pacific Islander, some other race, two or more races, and White. CBO used the composite category “some other race or multiracial” because of the small sizes of the groups of people who do not identify as Black or White. Each racial category can include people of any ethnicity, and either ethnic group can include people of any race or combination of races.

First Street Foundation’s flood model generated the property-level projections of risk that CBO used in this report. The agency considered a property as being exposed to flood risk if it has at least a 1 percent projected probability of flooding at depths of roughly 1 foot or more in a single year. A 1 percent annual probability of flooding is the same as a 26 percent chance of at least one such flood occurring over a period of 30 years.

Flood projections and information on National Flood Insurance Program policies are at the property level. Community characteristics, by contrast, reflect household-level survey responses and are reported at the block group level. Those data represent a profile of communities over the 2016–2020 period. For more details on terminology, see Appendix A; for details on the data and CBO’s method, see Appendix B.

Numbers in the text and figures may not add up to totals because of rounding.

The Congressional Budget Office previously examined the distribution of flood risk across communities in the United States with different economic and demographic characteristics.1 This report builds on that distributional analysis by examining how the prevalence of flood insurance policies purchased through the National Flood Insurance Program (NFIP)—those that cover structures, contents, or both—varies with the economic and demographic characteristics of communities at risk of flooding. In this report, the agency also examines how the prevalence of policies with premium discounts varies by characteristic.

To do that analysis, CBO used the results of a model that generates property-level flood projections for an intermediate global warming scenario in which greenhouse gas emissions remain roughly at their current levels until 2050 and then decline. It corresponds to an increase in the global average temperature of 1.9 degrees Celsius by 2100, relative to the average over the 1995–2014 period. The projections account for four sources of flooding—sea level rise, storm surge (water pushed ashore by winds of a storm, whether or not the storm makes landfall), rainfall, and overflowing rivers and streams—and for adaptive infrastructure, such as levees and wetland projects. Flood models generate projections for multiyear periods, and in this report, projections for the current period are centered on the year 2023.

CBO measured community characteristics using the Census Bureau’s economic and demographic data for the 2016–2020 period for block groups—the smallest areas for which such data are available. Those data reflect underlying information, which is not publicly available, about households (the people usually residing in a residence) or about householders (people in whose name a residence was owned or rented—or, if such a person was not present, any household member who was at least 15 years old).

CBO determined whether individual properties at risk of flooding were covered by an active NFIP policy on May 1, 2023, using data from the Federal Emergency Management Agency (FEMA). As of April 1, 2023, all policies had premiums determined under FEMA’s new method for assessing risk, which fully accounts for property-level flood risk. (The previous method assigned risk largely on the basis of flood maps that sorted areas into three general zones by level and type of risk. Those zones also served to designate areas at high risk of flooding as Special Flood Hazard Areas, or SFHAs. The designations have not been updated to incorporate the information FEMA uses in its new method for determining premiums.) Under the new method, roughly three-quarters of NFIP policyholders faced higher premiums.2 (For explanations of the terminology used in this report, see Appendix A; for details about the data and CBO’s method, see Appendix B.)

Policyholders may receive one or more premium discounts. Some are time limited; the most common of those discounts are legal caps on annual rate increases. (The Homeowner Flood Insurance Affordability Act of 2014 established a cap of 18 percent for primary residences, and the Biggert-Waters Flood Insurance Reform Act of 2012 established a cap of 25 percent for other types of properties, such as secondary residences and business properties.) Some policyholders also have time-limited discounts of 60 percent or 70 percent on the first $35,000 of building coverage and $10,000 of content coverage to account for changing information about flood risk. (In most cases, when those discounts expire, the legal caps on annual rate increases apply.)

Other discounts do not expire. Community Rating System (CRS) discounts range from 5 percent to 45 percent, depending on the extent to which a local government entity—a county or town, for example—implements floodplain management practices that exceed the NFIP’s minimum requirements. Most communities retain or increase their CRS discount over time. (For additional details about discounts, see Appendix B.)

Insurance for Properties at Risk of Flooding in the United States

First Street Foundation’s (FSF’s) data about flood risk in the United States represent roughly 145 million properties—residential, commercial, industrial, agricultural, or governmental, for instance—some of which are vacant. In this report, CBO considers those properties to be at risk of flooding if, in the FSF data, they have at least a 1 percent annual probability of experiencing a flood of a depth of at least 1 foot, which is equivalent to having about a one-in-four chance of experiencing at least one such flood over a period of 30 years. By that definition, about 9 percent of U.S. properties are at risk. Among properties at risk, about 92 percent were not covered by an active National Flood Insurance Program policy as of May 1, 2023.

‌‌Share of At-Risk Properties With and Without NFIP Coverage

Percent

Only about 8 percent of properties with at least a 1 percent annual probability of experiencing a flood of a depth of at least 1 foot were covered by an NFIP policy as of May 1, 2023. About two-thirds of those policies included a premium discount associated with the Community Rating System. CRS discounts do not expire. About one-fourth had a time-limited premium discount, and about one-tenth had no discount.

Many of the time-limited premium discounts expire after a policy’s first term of eligibility. When that happens, people are less likely to renew their NFIP policies. The Government Accountability Office estimates that in 2037, only 5 percent of policies will still be subject to the annual caps on premium rate increases.

Insurance for Properties at Risk of Flooding, by Special Flood Hazard Area Status

FEMA conveys flood risk information in part through maps that identify SFHAs, which are areas that have at least a 1 percent annual probability of flooding. However, many properties located outside SFHAs also have that level of risk because SFHA designations do not reflect updated flood modeling—notably, they do not account for flooding associated solely with rainfall. In fact, among the properties that CBO examined (those with at least a 1 percent annual probability of experiencing a flood of a depth of at least 1 foot), more are located outside SFHAs than inside SFHAs.

Within SFHAs, if NFIP coverage is available—that is, if the jurisdiction chooses to participate and meets the NFIP’s minimum floodplain management requirements—a property generally must be covered by flood insurance to be eligible for a mortgage from a federally regulated lender or for federal financial assistance. (Several additional factors, such as whether the property is insured through the NFIP or a private insurer and whether the property includes a structure, influence the share of at-risk properties in SFHAs with NFIP coverage; see Appendix B.)

‌Share of At-Risk Properties With and Without NFIP Coverage, by SFHA Status

Percent

In the FSF flood risk data, roughly two-thirds of properties with at least a 1 percent annual probability of experiencing a flood of a depth of at least one foot are located outside SFHAs.

Properties at risk in SFHAs were more likely to be covered by an NFIP policy than those located outside SFHAs were (18 percent and 4 percent, respectively).

‌Share of Covered At-Risk Properties With and Without Premium Discounts, by SFHA Status

Percent

The share of policyholders with a premium discount was about the same regardless of whether their property was located in an SFHA. However, the discount was more likely to be a CRS discount, and therefore non-expiring, for properties in SFHAs.

How CBO Measured Communities’ Flood Risk and Prevalence of Flood Insurance

CBO examined the distribution of flood risk across communities in the United States according to specific economic and demographic characteristics: median household income, the race of householders in a community, their ethnicity, the composition of their households, whether they used their property as a primary or secondary residence, whether they rented or owned their residence, and whether they lived in coastal or inland communities.

The agency used two methods to compare the risk across communities. In one, CBO grouped communities into quintiles (or fifths) according to the distribution of households with a given characteristic; in the other, the agency grouped communities according to the characteristic of the majority of householders. In all cases, although communities are described by characteristics of households or householders, risk is projected for properties, and CBO evaluated whether each property was covered by an NFIP policy. (The number of properties and the number of households are not necessarily the same because some properties are multifamily residential properties, some are not residential, and some are vacant.)

The economic and demographic information used in this analysis is available only at the community level, but the model generates flood projections at the property level. In addition, CBO considered whether individual properties were covered by a policy with the NFIP. Therefore, the properties at risk in any given quintile or majority group, as well as those with flood insurance, might or might not have been occupied by a householder who had a particular characteristic.

By Quintile

To construct the quintiles used in this report, CBO ordered all communities by their share of households with a specific characteristic and then divided the communities into five equal groups. The communities in the top quintile (one-fifth of all communities) have the largest shares of households with the characteristic. The communities in the bottom quintile have the smallest shares, and for some characteristics, communities in the lower quintiles have no households with that particular characteristic.

‌Calculating the Share of Properties at Risk of Flood by Quintile

Each small square represents a block group, or community, made up of between 240 and 1,200 households. The degree of shading in a square represents the share of households in the community with a specific characteristic.

After arraying the distribution of communities in order, CBO calculated the flood risk for each quintile.

By Majority

Many communities had no householders with a specific characteristic. When the number of such communities was sufficiently large, one or more of the quintiles had zero percent of households with that characteristic. In such cases, CBO supplemented its quintile analysis by grouping communities according to the characteristic of the majority of householders in a given community and examined the flood risk and prevalence of flood insurance in each group.

Insurance for Properties at Risk of Flooding, by Household Income

In this report, CBO distinguishes the economic characteristics of communities by comparing their median household income. Half of the households in a community have household income that is lower than the median, and half of the households have income that is higher. CBO found that the higher a community’s median household income, the smaller the share of properties at risk of flooding, the larger the share of at-risk properties covered by an NFIP policy, and the greater the prevalence of premium discounts among properties with coverage. Regardless of income, most policies had premium discounts.

‌Share of At-Risk Properties With and Without NFIP Coverage, by Quintile of Median Household Income

Percent

In communities in the bottom quintile for median household income, 11 percent of properties are at risk of flooding, but only 4 percent of those properties were covered by an NFIP policy. In the top quintile, 8 percent of properties are at risk, 14 percent of which were covered.

‌Share of Covered At-Risk Properties With and Without Premium Discounts, by Quintile of Median Household Income

Percent

Among properties that were covered, premium discounts were least prevalent in communities in the bottom quintile for median household income: 87 percent of those insured properties had a discount. By contrast, 92 percent of insured properties in the top quintile had one. Discounts were least likely to be CRS discounts in communities in the bottom quintile.

Insurance for Properties at Risk of Flooding, by Householders’ Race

CBO used the following racial groups in its analysis: Black, White, and some other race or multiracial. The last group consists of householders who identified as American Indian, Alaska Native, Asian, Native Hawaiian, Pacific Islander, some other race, or two or more races. Ethnicity was analyzed separately.

Race of the Majority of Householders in a Community

In 95 percent of all communities, a majority of the householders were in one of the racial groups used for this analysis. White householders made up the majority in 82 percent of communities, Black householders in 8 percent, and householders identifying as some other race or multiracial in 5 percent.

Properties at risk of flooding were least likely to be covered by an NFIP policy in communities in which the majority of the householders were some other race or multiracial. The share of properties with NFIP coverage was also relatively small in communities in which the majority of householders were Black. Policyholders were least likely to have premium discounts in communities in which most householders were some other race or multiracial.

‌Share of At-Risk Properties With and Without NFIP Coverage, by Majority of Householders’ Race

Percent

The share of at-risk properties that were covered by an NFIP policy ranged from 5 percent to 8 percent.

‌Share of Covered At-Risk Properties With and Without Premium Discounts, by Majority of Householders’ Race

Percent

In communities in which most householders were White, those in which most were Black, and those in which no racial group constituted a majority, at least 90 percent of policyholders received a premium discount. In communities in which most householders were some other race or multiracial, 85 percent had a premium discount, and it was least likely to be a CRS discount.

Race and Median Household Income

CBO also examined how the median household income of communities, grouped by race, affected how likely an at-risk property was to be covered by an NFIP policy. For any quintile of income, at-risk properties in communities in which the majority of the householders’ race fell in the “some other race or multiracial” category were least likely to be covered by an NFIP policy. Only in the top two income quintiles were at-risk properties less likely to be covered by an NFIP policy in communities in which the majority of householders were Black than in communities in which the majority of householders were White.

Quintile of Race

The communities with the smallest share of at-risk properties with NFIP coverage were those in the bottom quintile of the distribution of White householders (that is, those with the smallest percentages of White householders), those in the top quintile of the distribution of Black householders, and those in the top quintile of the distribution of householders who were some other race or multiracial. Most policyholders had premium discounts, but the shares varied across the distributions by quintile of race.

‌Share of At-Risk Properties With and Without NFIP Coverage, by Quintile of Householders’ Race

Percent

In communities in the top quintiles of the distributions of both householders who were Black and those who were some other race or multiracial (that is, communities with the largest percentages of those populations), the share of at-risk properties covered by an NFIP policy was 7 percent. (Within those quintiles, in communities with the largest shares of those householders, that share was smaller, at 5 percent.) By contrast, in other quintiles of those distributions, the share of covered properties was larger, at 8 percent or 9 percent.

‌Share of Covered At-Risk Properties With and Without Premium Discounts, by Quintile of Householders’ Race

Percent

Most policyholders in all quintiles of all three distributions had premium discounts. Discounts were most prevalent in the top quintile for White householders, at 92 percent, and least prevalent in the top quintile for householders who were some other race or multiracial, at 88 percent.

Insurance for Properties at Risk of Flooding, by Householders’ Ethnicity

CBO arranged householders into two ethnic groups: those who identified as Hispanic or Latino and those who did not identify as Hispanic or Latino. For simplicity, this report refers to those groups as Hispanic and not Hispanic, respectively. Because race is not tied to ethnicity in the data, the two ethnic groups include householders from all racial groups.

Ethnicity of the Majority of Householders in a Community

Householders who identified as Hispanic made up a majority of householders in 9 percent of communities; those who did not identify as Hispanic made up a majority of householders in 91 percent of communities. At-risk properties in communities in which the majority of householders were Hispanic were more likely to be covered by an NFIP policy. Policyholders were equally likely to have a premium discount whether the majority of householders in the community were Hispanic or not.

‌Share of At-Risk Properties With and Without NFIP Coverage, by Majority of Householders’ Ethnicity

Percent

Properties were more likely to have NFIP coverage in communities in which most householders were Hispanic.

‌Share of Covered At-Risk Properties With and Without Premium Discounts, by Majority of Householders’ Ethnicity

Percent

Policyholders were equally likely to receive premium discounts in communities in which most householders were Hispanic and in communities in which most were not. Those discounts were more likely to be CRS discounts in communities in which most householders were Hispanic.

Ethnicity and Median Household Income

CBO also examined how the median household income of communities, grouped by ethnicity, affected the relative prevalence of insurance for properties at risk of flooding. The agency found that the difference in the share of properties with an NFIP policy between communities in which a majority of householders identified as Hispanic and those in which a majority of householders did not was larger when median household income was higher.

Quintile of Ethnicity

Communities with a greater share of householders who were Hispanic generally had a larger share of at-risk properties covered by an NFIP policy. Discounts were common across all quintiles.

‌Share of At-Risk Properties With and Without NFIP Coverage, by Quintile of Householders’ Ethnicity

Percent

In communities in the bottom two quintiles of the distribution of householders who were Hispanic (which correspond to those in the top two quintiles for householders who were not), 7 percent of properties at risk were covered by an NFIP policy. In communities in other quintiles, that share was 10 percent.

‌Share of Covered At-Risk Properties With and Without Premium Discounts, by Quintile of Householders’ Ethnicity

Percent

The prevalence of premium discounts was relatively uniform across communities. Policyholders in communities in the bottom quintile for householders who were Hispanic (which corresponds to the communities in the top quintiles for householders who were not) were slightly less likely—by 1 percentage point to 2 percentage points—than policyholders in other quintiles to have a premium discount.

Insurance for Properties at Risk of Flooding, by Type of Residence

CBO analyzed the relationship between the prevalence of NFIP insurance and communities’ types of residences using two characteristics: whether residences were primary or secondary housing units and whether residences were rented or owned.

Primary or Secondary Residence

A primary residence serves as a householder’s only permanent residence; a secondary residence is used for seasonal, recreational, or occasional purposes. Most communities had no secondary residences. Only 1 percent of all communities in the United States were composed mostly of secondary residences, whereas 99 percent of communities were composed mostly of primary residences.

Communities in which the majority of dwellings were secondary residences have a larger share of properties at risk. Those communities also had a larger share of at-risk properties covered by an NFIP policy, and those policies were more likely to have a premium discount.

‌Share of At-Risk Properties With and Without NFIP Coverage Depending on Whether Most Residences in the Community Were Primary or Secondary

Percent

In communities in which most dwellings were secondary residences, 17 percent of properties at risk had an NFIP policy, roughly double the share in communities in which the majority of dwellings were primary residences.

‌Share of Covered At-Risk Properties With and Without Premium Discounts Depending on Whether Most Residences in the Community Were Primary or Secondary

Percent

In communities composed mostly of secondary residences, 94 percent of policies had premium discounts. In communities in which most dwellings were primary residences, that share was 90 percent.

In communities in which most dwellings were secondary residences, discounts were more likely to be CRS discounts, which do not expire.

Renter or Owner Households

Among properties at risk, the share covered by an NFIP policy was lowest in communities with the largest proportion of renters. Policies in those communities were also least likely to have a premium discount.

‌Share of At-Risk Properties With and Without NFIP Coverage, by Quintile of Renters

Percent

Properties in communities with the smallest shares of renters were most likely to be covered by an NFIP policy—10 percent of properties were covered. Properties in communities with a larger share of renters were 2 percentage points to 3 percentage points less likely to be covered.

‌Share of Covered At-Risk Properties With and Without Premium Discounts, by Quintile of Renters

Percent

In communities with the smallest shares of renters, 93 percent of NFIP policies had premium discounts. In communities with the largest shares, 87 percent of policyholders had premium discounts.

In both the bottom and top quintiles, premium discounts were equally likely to be CRS discounts.

Insurance for Properties at Risk of Flooding, by Household Composition

To examine how flood risk varies across communities by household composition, CBO grouped households in two ways: whether they included a resident age 65 or older—often a householder—and whether they included a child. CBO chose to examine those groupings because seniors and children are particularly vulnerable to disruptions caused by flooding in their community. The two categories are not mutually exclusive—that is, the presence of a senior in a household does not imply that there is not a child in the household.

Households With a Senior

Among properties at risk, communities with the largest proportion of households with a senior resident had the largest share of NFIP policies. Policies in those communities were also most likely to have a premium discount.

‌Share of At-Risk Properties With and Without NFIP Coverage, by Quintile of Households With a Senior

Percent

In communities in the top quintile for the share of households with a senior, 11 percent of properties at risk were covered. By contrast, in communities with a smaller share of households with a senior, the proportion of at-risk properties that were insured was smaller by at least 3 percentage points.

‌Share of Covered At-Risk Properties With and Without Premium Discounts, by Quintile of Households With a Senior

Percent

In communities in the top quintile for the share of households with a senior, 93 percent of NFIP policies had a premium discount, and they were most likely to have a CRS discount. In quintiles with a smaller share of households with a senior, no more than 90 percent of properties with policies had premium discounts.

Households With a Child

Among properties at risk, communities with the smallest proportion of households with a child had the largest share of NFIP policies. Policies in those communities were also most likely to have a premium discount.

‌Share of At-Risk Properties With and Without NFIP Coverage, by Quintile of Households With a Child

Percent

In communities in the bottom quintile for the share of households with a child, 13 percent of properties at risk had coverage, whereas in communities with a larger share of households with a child, 6 percent to 8 percent of properties at risk had NFIP coverage.

‌Share of Covered At-Risk Properties With and Without Premium Discounts, by Quintile of Households With a Child

Percent

In communities in the bottom quintile for the share of households with a child, 94 percent of NFIP policies had a premium discount. In quintiles with a larger share of households with a child, no more than 90 percent of policies had a premium discount.

Discounts were most likely to be CRS discounts in communities in the bottom quintile for the share of households with a child.

Insurance for Properties at Risk of Flooding, by Geographic Location

Overall, risk is more prevalent in coastal areas than in inland areas because the former have more exposure to potential sources of flooding. Among properties at risk, the share with an NFIP policy was higher in coastal communities than in inland communities. In part, that is because FEMA’s flood maps, and the associated SFHAs (where flood insurance is required under certain circumstances), do not reflect updated flood modeling and do not account for flooding associated solely with rainfall. Policies in coastal communities were also more likely to have a premium discount.

‌Share of Coastal and Inland Properties at Risk With and Without NFIP Coverage

Percent

In coastal communities, 23 percent of properties at risk were covered by an NFIP policy, whereas only 4 percent of properties at risk in inland communities were covered by an NFIP policy.

‌Share of Covered At-Risk Coastal and Inland Properties With and Without Premium Discounts

Percent

In coastal communities, 95 percent of policies had premium discounts. By contrast, 84 percent of policies for properties in inland communities had premium discounts. Among policies with discounts, those for properties in coastal communities were more likely to have CRS discounts.

1. Congressional Budget Office, Communities at Risk of Flooding (September 2023), www.cbo.gov/publication/58953.

2. Federal Emergency Management Agency, “NFIP’s Pricing Approach” (accessed February 20, 2024), www.fema.gov/flood-insurance/risk-rating.

Appendix A: Glossary

American Community Survey (ACS). A nationwide survey used to collect demographic, social, housing, and economic information every year by contacting people at a sample of addresses. The ACS replaced the decennial census long form. It currently produces one-year and five-year estimates for small areas, including census tracts and block groups.

block group. A statistical subdivision of a census tract. It generally contains between 600 and 3,000 people living in 240 to 1,200 housing units. A block group is the smallest geographic unit for which the Census Bureau tabulates sample data. Block groups are created to reflect largely homogeneous economic, demographic, and housing characteristics.

census tract. A statistical subdivision of a county. Tract boundaries normally follow visible features such as rivers and roads, but in some cases, they may follow legal boundaries and other features that are not visible. Census tracts generally contain between 1,200 and 8,000 people living in 480 to 3,200 housing units.

community. A census block group identified in the 2020 Census.

coastal community. A community that meets two criteria: One, the community is located in a county on the country’s saltwater coast; and two, it consists partly of water. Thus, coastal communities in this analysis either lie on one of the country’s saltwater coasts or are located in a county on the coast and contain at least part of another body of water, such as a river or a lake. (See Appendix B, “Geographic Location of Communities.”)

dwelling. A residential property.

flood risk. At least a 1 percent projected annual probability of experiencing a flood of a depth of roughly 1 foot or more. A 1 percent annual chance of flooding is the same as a 26 percent chance of experiencing at least one such flood over a period of 30 years.

household. Consists of all the people who occupy a housing unit as their usual place of residence.

householder. A person in whose name a housing unit is owned or rented. (If there is no such person present, any household member who is at least 15 years old can serve as the householder.)

housing unit. Any residential space that can be occupied as separate living quarters with access outside, either directly or through a common hall. A house, an apartment, a mobile home or trailer, a group of rooms, or a single room may be considered a housing unit. However, group living arrangements that are owned or managed by an entity that provides housing or services (such as college residence halls, skilled nursing facilities, and military barracks) are not considered housing units.

prevalence of flood risk. The percentage of properties facing at least a 1 percent annual probability of experiencing a flood of a depth of roughly 1 foot or more.

primary residence. A housing unit occupied by people who consider it their sole usual place of residence.

property. A parcel of land that may be used for residential, commercial, industrial, agricultural, governmental, educational, religious, or other purposes. (The data do not contain any information about types of structures, and structures are not necessarily present on each property.)

quintile. One of five equal-sized groups into which a population can be divided according to the distribution of values of a particular variable.

race and ethnicity. The Congressional Budget Office grouped householders on the basis of their self-identification as recorded in the ACS’s five-year estimates. For this analysis, the racial groups (Black, White, and some other race or multiracial) and the ethnic groups (Hispanic and not Hispanic) were constructed using the estimates for the 2016–2020 period. The racial categories specified in those estimates are American Indian and Alaska Native, Asian, Black or African American, Native Hawaiian and Pacific Islander, some other race, two or more races, and White. CBO used the composite group “some other race or multiracial” because of the small sizes of the groups of people who did not identify as Black or White. The ethnic groups specified in the ACS’s estimates were Hispanic or Latino and neither Hispanic nor Latino. For simplicity, this report refers to those groups as Hispanic and not Hispanic, respectively. Each racial category can include people of any ethnicity, and either ethnic group can include people of any race or combination of races.

secondary residence. A secondary residence is a housing unit for seasonal, recreational, or occasional use that, at the time of the census interview, is either vacant or occupied by people who have a primary residence elsewhere.

Appendix B: Data and Methods

The Congressional Budget Office relied on data and projections from the Census Bureau, First Street Foundation (FSF), and the Federal Emergency Management Agency (FEMA) for its analysis in this report. This appendix describes those data as well as CBO’s methods for assessing how the adoption of National Flood Insurance Program (NFIP) policies within communities at risk of flooding varies with community characteristics. The flood risk projections were generated by FSF for one of the warming scenarios identified by the Intergovernmental Panel on Climate Change (IPCC).

Community Characteristics From the Census Bureau

The Census Bureau conducts a full census every 10 years, most recently in 2020, but it also contacts people at a sample of addresses annually to collect information about people and their households through the American Community Survey (ACS). From the survey responses, the Census Bureau produces five-year estimates of demographic, social, housing, and economic characteristics for small areas known as block groups, or communities, each of which contains between 600 and 3,000 people living in 240 to 1,200 housing units.

CBO used the ACS’s five-year estimates for the 2016–2020 period to determine the characteristics of communities.1 Those estimates apply to 242,335 block groups identified in the 2020 census. To protect privacy, however, the Census Bureau suppresses data on characteristics if a block group’s population is too small. Data on median household income were withheld for 6 percent of block groups, and data on the other characteristics examined in this report were withheld for 1 percent.

The most detailed information publicly available on community characteristics is measured at the block-group level. Within a given block group, CBO’s analysis cannot account for the specific locations of households with shared characteristics, so the analysis cannot capture potentially relevant property-level variation in the adoption of NFIP policies among properties at risk. Consider this example: In communities in which the majority of the householders are White, the share of properties at risk that are insured is larger than the share of properties at risk that are insured in communities in which a different racial group constitutes a majority of the householders. It does not necessarily follow, however, that in communities in which the majority of householders are White, it is White householders, among those at risk, who are most likely to have insurance.

CBO’s analysis also does not capture potentially relevant variation in flood risk or in the prevalence of insurance associated with structure type, which may be correlated with household characteristics. For example, the flooding of a multistory apartment building would primarily affect units on the ground floor, and the data report whether there are insurance policies associated with that multistory apartment building but not which units have insurance.

Geographic Location of Communities

In addition to examining flood risk on the basis of economic and demographic characteristics, CBO also looked at how risk differed depending on a community’s location. The agency used two criteria to determine whether a community was coastal: whether it was located in a coastal county and whether the specific block group contained any water, whether ocean, gulf, lake, or other body of water. That is, all coastal communities are located in coastal counties, but not all communities in coastal counties are considered coastal communities in this report; likewise, all coastal communities located on one of the nation’s saltwater coastlines are considered coastal communities, but not all coastal communities are located on one of those coastlines. CBO used the Census Bureau’s publications to identify coastal counties and the Census Bureau’s TIGERweb State-Based Data Files to identify block groups that contain at least part of a body of water.2 Accordingly, 10 percent of U.S. block groups are coastal; the remaining 90 percent are inland.

CBO’s analysis may understate the differences in community risk between coastal and inland areas. CBO’s definition of a coastal community includes some block groups that, although in coastal counties, do not lie on the coast. Those block groups (which are more likely to be in states with relatively large counties) may be less exposed to certain types of flooding, such as that associated with rising sea levels and storm surges.

The majority of coastal properties and the majority of inland properties have no projected risk of flooding (see Table B-1). Roughly one-quarter of all the properties in coastal areas have a projected risk of flooding greater than zero, and the same is true for roughly one-tenth of all the properties in inland areas. In both areas, most properties with a projected nonzero risk are projected to have at least a 1 percent annual probability of experiencing a flood of a depth of at least 30 centimeters (about 1 foot; see Table B-2). Those are the properties that CBO labels at risk in this analysis.

Table B-1.

Projected Depth of a Flood With at Least a 1 Percent Annual Chance of Occurring

Centimeters

Notes

Data source: Congressional Budget Office. See www.cbo.gov/publication/60042#data.

Table B-2.

Projected Depth of a Flood With at Least a 1 Percent Annual Chance of Occurring for Properties With a Risk of Flooding Greater Than Zero

Centimeters

Notes

Data source: Congressional Budget Office. See www.cbo.gov/publication/60042#data.

Projections of Flood Risk From First Street Foundation

Projections of flood risk are uncertain, and CBO aims to provide projections that are in the middle of a range of possible outcomes. The IPCC identified several warming scenarios known as shared socioeconomic pathways (SSPs) in its sixth assessment report.3 FSF generated flood risk projections for one of those scenarios, SSP2-4.5, which is an intermediate warming scenario. In that scenario, emissions remain around current levels until 2050 and then decline but do not reach net zero by 2100, yielding a global average temperature increase of 1.9 degrees Celsius over the average for the 1995–2014 period. (Impacts of a warmer climate occur in every scenario modeled by the IPCC, including one with zero net global emissions by 2050.) FSF’s projections that CBO used in its analysis represent the middle of the distribution of outcomes from climate models used by the IPCC to examine SSP2-4.5.

FSF combines data on historical floods from 1950 to 2014 with climate modeling from the IPCC to project flood depths for 145.4 million properties, each of which is matched to its block group in the census. FSF’s data cover all properties in 99 percent of the census’s block groups.

Projected flood depths incorporate the effects of existing adaptive infrastructure, but the data that such large-scale flood models can use to account for the existence and condition of adaptive infrastructure are limited. FSF’s modeling incorporates data from the National Levee Database and the National Inventory of Dams, both maintained by the Army Corps of Engineers, as well as a supplementary database of adaptive infrastructure developed by FSF.4 That supplementary database, which also accounts for the useful life of the infrastructure, draws on varied sources, including FEMA’s flood insurance studies, states’ coastal zone management authorities, and minutes from public meetings. The data include structures (such as levees, dams, and hardened ditches) and nature-based projects designed to slow flood waters (such as wetland projects).5 Still, the infrastructure data are incomplete.

For this report, CBO relied on the third version of FSF’s flood risk projections, which incorporates information about recent increases in heavy precipitation events and also reflects the most recent set of climate models and scenarios from the IPCC.6 Improvements in property identification account for new development, and the property data in the new version include 10 small counties that were not represented in the previous version’s projections.

National Flood Insurance Policies From the Federal Emergency Management Agency

Information on all active NFIP policies, the address associated with each policy, and applicable discounts is from FEMA’s PIVOT system.7 Data on active NFIP policies as of May 1, 2023, were the most recent available in which all NFIP premiums had been determined under FEMA’s new method for assessing a property’s risk, called Risk Rating 2.0, which fully accounts for property-level flood risk.8

Of the SFHA properties with at least a 1 percent probability of flooding at depths of roughly one foot or more in the FSF data that CBO used for its analysis, several factors determine the share that is covered by NFIP policies. First, that share is influenced by the number of those properties that include a structure and, of those, how many are subject to the mandatory purchase requirement. SFHA properties are subject to the requirement if they are in a community that participates in the NFIP (currently, about 10 percent of jurisdictions with an SFHA do not participate in the NFIP) and if the property owner received federal financial assistance. (For example, homeowners or business owners receive such assistance when they have a government-backed loan on their building, which characterizes most private mortgages—although roughly 40 percent of homes are not mortgaged.)9 Second, that share is affected by how many property owners comply with the requirement by obtaining non-NFIP flood insurance from a private insurer. (The NFIP is more likely to insure residential properties serving as primary residences, whereas private insurers are more likely to provide coverage for properties that host other types of structures.) Finally, the share of SFHA properties at risk that have NFIP coverage is also influenced by the rate of compliance with the mandatory purchase requirement.10

Lenders are responsible for enforcing the requirement. Regulators have levied civil monetary penalties on lenders that were in substantial violation of the requirement. Separately, the contracts between government-sponsored enterprises (Fannie Mae and Freddie Mac) and loan sellers and servicers lay out consequences for noncompliance, which can include reimbursement of the federal guarantor for uninsured losses caused by flooding or repurchase of the loans by the seller or servicer.11

Moreover, in the wake of a major disaster in an SFHA in a community that participates in the NFIP, FEMA offers a three-year insurance policy (standard NFIP policies have a one-year term) to survivors who are approved for disaster assistance and who meet other eligibility requirements. The cost of the policy is covered by the funding for disaster assistance. If recipients do not maintain flood insurance, they may be ineligible for certain types of disaster assistance in the future.

Many policyholders receive one or more premium discounts. The most common one is the Community Rating System (CRS) discount, which does not expire. CRS discounts range from 5 percent to 45 percent; the precise discount depends on the extent to which a local government entity (a county, city, or town, for example) executes floodplain management practices that exceed the minimum NFIP requirements. Although communities can stop participating in the CRS, few do so. At least 90 percent of the communities that participated in the CRS in or after 1999 continued to participate through 2021.12 Among communities with continuous participation, 32 percent maintained their CRS discount and 52 percent saw it increase over time because of their floodplain management practices. Thirteen percent experienced some variation in their discount percentage over the period but by 2021 had either regained or exceeded their original discount. A small share, 3 percent, either experienced variation and by 2021 had a discount percentage that was lower than their original one or saw their discount lowered because of their floodplain management practices.13

Among time-limited discounts, legal caps on annual rate increases are the most common. Increases are capped at 18 percent for primary residences and 25 percent for other types of properties. Some policyholders have other time-limited discounts; those usually take the form of a 60 percent or 70 percent discount applied to the first $35,000 of building coverage and $10,000 of content coverage. When those discounts expire, the legal caps on annual rate increases usually apply (see Table B-3).

Table B-3.

Time-Limited Premium Discounts in the National Flood Insurance Program

Notes

Data source: Congressional Budget Office, using information from the Federal Emergency Management Agency.

FEMA = Federal Emergency Management Agency; FIRM = Flood Insurance Rate Map; NFIP = National Flood Insurance Program; SFHA = special flood hazard area.

SFHAs are areas with at least a 1 percent annual chance of flooding.

How certain discounts are applied to a policy premium depends on how close it is to one that would fully account for the property’s flood risk. For premiums on existing policies, the NFIP applies any applicable CRS discount before it applies the legal cap on the annual rate increase. Accordingly, the annual price increase for a policyholder with a CRS discount who also benefits from the cap is determined solely by the cap until the premium reaches a level at which the cap no longer applies.14

Matching National Flood Insurance Policies and Flood Risk Projections by Property

For its analysis, CBO determined whether individual properties were at risk, whether they were covered by an NFIP policy as of May 1, 2023, and whether they were in an SFHA. To do so, the agency matched NFIP policies from the FEMA data to properties from the FSF data.

FEMA processed its data with the help of a contractor who transformed each recorded NFIP policy address to a standardized address and then matched that standardized address to a set of geographic coordinates. For each policy, a “match code” characterized the quality of the match. Using Google Maps, CBO verified that policies with certain match codes were usually associated with a standardized address and geographic coordinates that both pointed to the same location. Policies with other match codes were more often associated with a standardized address and geographic coordinates that pointed to locations over 250 feet apart. Those less reliable match codes accounted for 5.3 percent of policies; CBO did not include them in its analysis.

To match NFIP policies in the FEMA data to their corresponding properties in the FSF data, CBO applied an algorithm to the lists of geographic coordinates in the two data sets. That algorithm matched each NFIP policy to the geographically closest FSF property.15 If that FSF property was at least 20 meters (or about 65 feet) away, CBO assumed that the NFIP policy was not matched to the correct property and that the property was uninsured. That was the case for 6.3 percent of the matched policies. If multiple NFIP policies matched to the same property, CBO counted that property as insured, and if at least one of those policies had a premium discount, the agency counted that property as receiving as a discount. Using the matched FEMA and FSF data, CBO designated each of the roughly 145 million properties in the United States as not having an NFIP policy, having an NFIP policy with a premium discount, or having an NFIP policy with no premium discount.

The NFIP policies that CBO dropped from the data—either because of their match code or because they were matched to a property over 20 meters away—had a state-level distribution similar to that of all NFIP policies. However, the dropped policies were 12 percent more likely to be in an SFHA compared with the full sample of policies.

Sources of Uncertainty

Uncertainties in flood modeling (other than those uncertainties related to climate modeling more generally) are primarily associated with data about terrain because flood patterns are mostly influenced by topography. Even large-scale data sets that describe terrain with relatively small errors in elevation can lead to mispredictions of whether a specific area will flood.16 Moreover, the resolution of the flood model determines the size of the square areas of physical land represented in the modeling grid. FSF’s model resolution is 30 meters (about 98 feet) and, where relevant local topographic data are available, a downscaling algorithm is used to resample the model output to a 3-meter resolution.17 Many topographic features, such as curbs and walls, are even smaller than that. The resolution of large-scale flood models makes it difficult to represent complex flooding patterns, and complex patterns are more common in urban areas.18

Additional uncertainties are associated with modeling stream flows that are different from those regularly experienced. For instance, some uncertainty arises from the initial assumption (necessary in a large-scale model) that river gauge records stretching back at least 30 years are applicable to a river in its current state. Uncertainties are also associated with the parameter values that influence how the models simulate the propagation of water through channels and over floodplains.19

1. Census Bureau, “2016–2020 ACS 5-Year Estimates” (March 31, 2022), https://tinyurl.com/msk3hw3k.

2. For more information about coastal counties, see Steven G. Wilson and Thomas R. Fischetti, “Coastline Population Trends in the United States: 1960 to 2008,” Report P25-1139 (Census Bureau, May 2010), https://tinyurl.com/3dcwbf3x; Census Bureau, “Substantial Changes to Counties and County Equivalent Entities: 1970–Present,” https://tinyurl.com/386k7mvf (2010) and https://tinyurl.com/2mu2ywj5 (2000). For block groups that include water, see Census Bureau, “TIGERweb State-Based Data Files” (accessed August 16, 2022), https://tinyurl.com/3af37ywe.

3. For details about specific scenarios, see H. Lee and J. Romero, eds., Climate Change 2023: Synthesis Report (Intergovernmental Panel on Climate Change, 2023), www.ipcc.ch/report/ar6/syr/.

4. The National Levee Database, in particular, is known to be incomplete. See American Society of Civil Engineers, A Comprehensive Assessment of America’s Infrastructure (2021), “Levees,” https://tinyurl.com/3suw2brd.

5. First Street Foundation, First Street Foundation Flood Model (FSF-FM): Technical Methodology Documentation, version 3.0 (July 31, 2023), https://tinyurl.com/4h7mjdpr (PDF).

6. First Street Foundation, “Property Level Flood Risk Statistics,” version 3.0 (July 31, 2023). CBO’s previous report on the distribution of flood risk across communities in the United States relied on an earlier version of FSF’s flood risk projections. See Congressional Budget Office, Communities at Risk of Flooding (September 2023), www.cbo.gov/publication/58953, and First Street Foundation, “First Street Foundation Property Level Flood Risk Statistics,” version 2.0 (April 13, 2022), https://zenodo.org/record/6459076.

7. The only discount CBO did not account for in this analysis is the 5 percent mitigation discount for elevating certain covered machinery and equipment and appliances that service a building to a level above the first floor.

8. Premiums are determined by the location of the property (including distance from flooding sources and elevation relative to the surrounding area), the characteristics of the building (including the type of construction, such as whether the walls are masonry or wood frame), the replacement cost value of the building, and the amount of coverage requested. See Federal Emergency Management Agency, “Rate Explanation Guide,” https://tinyurl.com/fjcwed6r (PDF), and National Flood Insurance Program: Risk Rating 2.0 Methodology and Data Sources (January 2022), https://tinyurl.com/ydnwsp8e (PDF).

9. Carolyn Kousky and others, “Flood Risk and the U.S. Housing Market,” Journal of Housing Research, vol. 29, no. S1, (November 2020), https://doi.org/10.1080/10527001.2020.1836915.

10. The 6.2 percent of properties that CBO dropped from its sample—because the match between FEMA data and census data was inaccurate—were 12 percent more likely to be in an SFHA compared with the full sample of policies.

11. Fannie Mae and Freddie Mac delegate underwriting authority to loan originators subject to a legal agreement that the loans meet specified criteria. The government-sponsored enterprises can require sellers to buy back the loans if the criteria are not met, but they do not always do so. During the financial crisis of the late 2000s, for example, the percentage of the sample of performing loans (those loans for which borrowers were making payments on time) reviewed by Freddie Mac and found to be in breach of the agreement rose from 10 percent for mortgages originated in 2005 to 23 percent for mortgages originated in 2008, but Freddie Mac required very few borrowers to repurchase those performing loans. See Financial Crisis Inquiry Commission, The Financial Crisis Inquiry Report: Final Report of the National Commission on the Causes of the Financial and Economic Crisis in the United States (January 2011), www.govinfo.gov/app/details/GPO-FCIC.

12. Most of the communities (54 percent) that participated in the CRS over the period covered by the data were already in the program in 1999. Before 1999, the process by which community floodplain management practices were evaluated for CRS ratings was relatively less stable.

13. Jesse Gourevitch, email message to author, providing data from Bill Lesser (February 27, 2024).

14. Federal Emergency Management Agency, “Community Rating System Discount Guide” (July 2023), https://tinyurl.com/fwj5wzfp (PDF).

15. CBO used the “scikit-learn” Python module to apply a ball-tree nearest-neighbors algorithm; see https://scikit-learn.org/stable/modules/neighbors.html for documentation.

16. Oliver E. J. Wing and others, “Estimates of Present and Future Flood Risk in the Conterminous United States,” supplementary data, “5.1: Uncertainty in Hazard Estimation,” Environmental Research Letters, vol. 13, no. 3 (February 2018), https://iopscience.iop.org/article/10.1088/1748-9326/aaac65.

17. Paul D. Bates and others, “Combined Modeling of U.S. Fluvial, Pluvial, and Coastal Flood Hazard Under Current and Future Climates,” Water Resources Research, vol. 57, no. 2 (February 2021), https://doi.org/10.1029/2020WR028673; First Street Foundation, First Street Foundation Flood Model (FSF-FM): Technical Methodology Documentation, version 3.0 (July 31, 2023), https://tinyurl.com/54k5ezjr (PDF).

18. Oliver E. J. Wing and others, “Validation of a 30 m Resolution Flood Hazard Model of the Coterminous United States,” Water Resources Research, vol. 53, no. 9 (September 2017), section 3.5, https://doi.org/10.1002/2017WR020917; Brett F. Sanders and others, “Large and Inequitable Flood Risks in Los Angeles, California,” Nature Sustainability, vol. 6, no. 1 (January 2023), www.nature.com/articles/s41893-022-00977-7.

19. Oliver E. J. Wing and others, “Estimates of Present and Future Flood Risk in the Conterminous United States,” Environmental Research Letters, vol. 13, no. 3 (February 2018), supplementary data, “5.1: Uncertainty in Hazard Estimation,” https://iopscience.iop.org/article/10.1088/1748-9326/aaac65; and Paul D. Bates and others, “Combined Modeling of U.S. Fluvial, Pluvial, and Coastal Flood Hazard Under Current and Future Climates,” Water Resources Research, vol. 57, no. 2 (February 2021), https://doi.org/10.1029/2020WR028673.

Appendix C: Data Sources for Figures

Share of At-Risk Properties With and Without NFIP Coverage

Congressional Budget Office, using data from First Street Foundation, “Property Level Flood Risk Statistics,” version 3.0 (July 31, 2023), and data from the Federal Emergency Management Agency.

Share of At-Risk Properties With and Without NFIP Coverage, by SFHA Status

Congressional Budget Office, using data from First Street Foundation and the Federal Emergency Management Agency.

Share of Covered At-Risk Properties With and Without Premium Discounts, by SFHA Status

Congressional Budget Office, using data from First Street Foundation and the Federal Emergency Management Agency.

Calculating the Share of Properties at Risk of Flood by Quintile

Congressional Budget Office.

Share of At-Risk Properties With and Without NFIP Coverage, by Quintile of Median Household Income

Congressional Budget Office, using data from Census Bureau, “American Community Survey 5-Year Data (2016–2020)” (March 31, 2022), https://tinyurl.com/msk3hw3k; First Street Foundation; and the Federal Emergency Management Agency.