Does National Flood Insurance Program Participation Induce Housing Development?

• Author: Carolyn A. Dehring,David L. Eckles,Mark J. Browne,William D. Lastrapes
• Date: 01 December 2019
• DOI: http://doi.org/10.1111/jori.12240
• Published date: 01 December 2019

©2018 The Journal of Risk and Insurance. Vol.XX, No. XX, 1–25 (2018).

DOI: 10.1111/jori.12240

Does National Flood Insurance Program

Participation Induce Housing Development?

Mark J. Browne

Carolyn A. Dehring†

David L. Eckles

William D. Lastrapes

Abstract

We estimate the effects on county-wide housing development in Florida of

community-level participation in the National Flood Insurance Program us-

ing panel data methods and observed variation in the timing of the decision

to join the program. Wefind that program participation “increased” housing

starts and permits in noncoastal counties, while coastal counties with av-

erage flood zone acreage saw significant “declines” in housing activity. We

conjecture that loss-mitigating regulatory costs, an obligation of obtaining

insurance, might explain these findings.

Introduction

Large-scale government insurance programs, of which there are many in the United

States and other developed countries, can have consequences that are undesirable,

albeit predictable. By insuring against loss, all forms of insurance can lead to moral

hazard and inefficient increases in risky behavior. But government-sponsored insur-

ance programs—especially those with below-cost premiums determined through the

political process under interest-group pressure,and ultimately subsidized by taxpay-

ers can distort behavior and increase risk more than private insurance.1Indeed, flood

Mark J. Browne is the Robert Clements Distinguished Chair in Insurance at the School of Risk

Management, Tobin College of Business, St. John’s University. Browne can be contacted via e-

mail: brownem1@stjohns.edu. Carolyn A. Dehring was at the Department of Insurance, Legal

Studies and Real Estate, Terry College of Business, University of Georgia.David L. Eckles is at

the Department of Insurance, Legal Studies and Real Estate, TerryCollege of Business, Univer-

sity of Georgia. Eckles can be contacted via e-mail: deckles@uga.edu. William D. Lastrapes is

at the Department of Economics, Terry College of Business, University of Georgia. Lastrapes

can be contacted via e-mail: last@uga.edu.

†Our friend and colleague, Carolyn Dehring, passed away March 12, 2015, during the latter

stages of this research. Wededicate this article to her memory.

1Kousky, Luttmer, and Zeckhauser (2006) shows that governmental programs do not always

result in a socially optimal solution. Originally,insured properties were explicitly subsidized

in the NFIP. More recently, attempts to institute rates closer to actuarially fair rates have met

with considerable political opposition (e.g., the Biggert–Waters Act).

1

835

835

Vol. 86, No. 4, 835–859 (2019).

2The Journal of Risk and Insurance

insurance would not be immune from this effect, and has been purported to lead

to adverse consequences (e.g., Cross, 1989). The extent to which flood insurance pro-

vided by the government increases risk and affects economic decisions is an empirical

question, and one that we examine in this article.

We estimate the effect of community-level participation in the National Flood Insur-

ance Program (NFIP), which allows owners of property in member communities to

insure their property from flooding, on local housing development. It is well known

that flood insurance premiums offered through the program have often underesti-

mated the expected property losses from flooding.2If the opportunity to buy flood

insurance at low rates has induced development—put more property in harm’s way—

we would expect the data to show that housing activity in participating communities

occurs at a faster rate than in nonparticipating communities, all else the same.

Our study employs data for housing activity and community-level NFIP membership

in the state of Florida over the period 1975–1989, during which time almost all com-

munities in the state joined the NFIP.3Although the NFIP covers the nation as a whole,

Florida accounts for a large portion of the nation’s active flood insurance policies and

total property exposure.4Given the geographic characteristics of the state, with its ex-

tensive coastal regions and overall vulnerability to severe weather and flooding, how

the flood insurance program has affected housing in Florida is particularly relevant

for policymakers.

We runpanel data regressions to explain how variation in housing starts and permits

is related to variation across years and counties in community-level membership in

the NFIP. In the “The NFIP” section, we discuss the history of the NFIP and the way

the program is operated. Crucially for our study, the flood mapping of communities

before they are eligible to participate in the “Regular Program” was done on a sched-

ule determined by the NFIP. Our evidence suggests that the NFIP scheduled mapping

dates based on the likelihood of an area being flooded, not on the rate of development

in the area. Thus, we treat the variation in county-wide intensity of membership as

exogenous. In this sense, the NFIP experience in Florida provides a “natural experi-

ment” for examining how housing markets respond to the availability of subsidized

government flood insurance.The “Empirical Model” section describes the empirical

model we use to exploit this natural experiment.

We are aware of only two studies that have directly examined the relationship be-

tween the NFIP and housing development. The first is a report from the U.S. Govern-

ment Accountability Office (GAO), which examined the rate of development and the

2Shavell (2014) shows that some governmental subsidies may be appropriate for “large risks”

like floods.

3We stop in 1989 to provide a cleaner test of development. This shorter time period allows us

to limit possibly confounding time effects and effects of the Community Rating System. When

we extend our data to 1998 (our last year of available data) and control for the effects of the

Community Rating System, our results remain largely unchanged.

4As of January 2015, 37 percent of such policies and exposure were in Florida;

http://bsa.nfipstat.fema.gov/reports/1011.htm

2The Journal of Risk and Insurance

836