Recent research paper on market flood insurance is misleading at best

"This country has come to feel the same when Congress is in session as when the baby gets hold of a hammer." - Will Rogers

Someone handed me a little factsheet about the great promise of private market flood insurance. You can find it at: https://www.milliman.com/insight/2017/Could-private-flood-insurance-be-cheaper-than-the-NFIP/ 

Let me start by saying that I’m a big supporter of free markets where they work. In the long history of the western world, no nation has prospered in the absence of a dominant market economy. I’m confident that the market can give us a better flood insurance outcome as long as private participation is structured consistent with our overall social goals regarding flood insurance.

The little paper displays some impressive graphs showing that large proportions of ALL HOMES might enjoy cheaper premiums with a private flood insurance market. The actuaries ran a flood risk model to develop actuarial premiums for flood insurance. They have these models just as the government does. They compared their hypothetical, computed rates to FEMA NFIP rates for ALL HOMES in three states, Florida, Texas and Louisiana, and counted the homes that would enjoy cheaper premiums under the private alternative. They did get some impressive results over ALL HOMES. 

Here's a representative statement: “Overall, across the three states, we found that a majority of single family homes could see cheaper premiums with private insurance than with NFIP. Based on our estimates, this would hold for 77% of the single family homes in Florida, 69% in Louisiana, and 92% in Texas.”

The little paper is laced with bar graphs making melodious variations on this theme. ALL SWEET HOMES.

So, why didn’t these people give us the premium comparisons for the NFIP pool of homes. There’s a good answer and they tell us in the paper. NFIP does not report the locations of the NFIP policy holders so the actuaries don’t know which homes in their model should be identified as NFIP homes.

It’s not that we should fault the idea here. Private insurance could well turn in some lower premiums than the government program which is structured to share the costs according to affordability constraints. The real problem is that the basis for comparison – ALL HOMES -- gives us some numbers that in the bright light of reality are much less meaningful than they look and at worst are misleading. 

Let’s see what I’m saying with a simple example. Suppose I look at a fourth, imaginary state. Let’s call it FLATX that has 100 homes (ALL HOMES). Suppose that 70 homes don’t have NFIP and that 30 do. I run the fancy actuarial model, and I find that of all 70 homes not in NFIP would get “cheaper premiums.” Suppose also that my model says that 3 (= 10%) of the NFIP homes get cheaper computed “private” premiums. So I conclude that 73% of ALL HOMES would get cheaper premiums. But really only 10% of the NFIP participating homes get cheaper premiums. So I report that 73% of ALL HOMES could enjoy cheaper premiums. Actually, in the little paper, the actuaries didn’t know which homes were not in NFIP and which were. But using the 73% number in my example would be misleading in terms of “representativeness” if I am more interested in the impact on the NFIP pool that has to buy the insurance by mandate. 

My critics would justly assert, “You don’t know that about these proportions of the homes.” Granted! They would be absolutely right, but they DON’T know just the same amount that I don’t. We’re equally ignorant. We’re running the risk here of exaggerating the case for private flood insurance in FLATX. We probably ought to be more interested in premium impacts to the NFIP population in FLATX than in the entire state. That proportion would be 10% in my example – not 73%. The 73% number might have some bearing on wanting to attract lots of people to buy flood insurance who are not buying it now. 

There’s also a significant conceptual problem with using premium reductions as a surrogate for benefits in this presentation. The reason is that for those households not participating in the NFIP program, the value of flood insurance is less than the FEMA premium. Otherwise, under the assumption of rational households, they would participate. Thus looking at a premium reduction as FEMA premium minus actuary estimated premium overstates the benefit of private insurance to those not now participating. The same is not true for those who do participate. Making “cheaper premium” statements for both the NFIP and non-NFIP pools together is mixing apples and oranges if one intends to communicate about the prospective benefits of private insurance. It’s conceptually wrong. In understanding this, the non-technical reader’s intuition should fix on the idea that the person not buying the NFIP product values it at less than the FEMA premium or he would buy it now. [i] 

So what do we know. Not much. We know that someone can run a model and produce private flood insurance rates that appear to be competitive with NFIP but these percentages (like 92% savings in TX) are misleading and may distort the desirability of the private insurance. We can put that observation beside the statements in reports from GAO and others over the years that NFIP charges “below market premiums.” The idea of these actuarial rates being computed lower than NFIP rates also clashes with the perennial reform concerns that we need to increase NFIP rates to actuarial levels to enhance the fiscal soundness of the program. And we’re left really flummoxed in consideration of the prospect that we have seen some decades of huge deficits in the NFIP pool but the private market with lower rates than NFIP as computed in this new actuarial flood model can make money. I grew up watching the “Baseball Game of the Week” with Dizzy Dean and Pee Wee Reese. As old Diz used to say, “Who’d have thunk it?” These new ideas about making money in flood insurance sure are at odds with everything we’ve been taught for the last couple of decades.

I’m also troubled by the statement that profitable private insurance is possible with a 35% loss ratio. The last FEMA Actuarial Rate Review I can find (https://www.fema.gov/media-library-data/20130726-1809-25045-6893/actuarial_rate_review2011.pdf) reported a loss ratio for NFIP of 43.8%. So, this little actuarial paper says we can have a lower loss ratio on lower premiums and still make money. I would like more information. Perhaps they hot new private insurance has lower benefits?

I conclude we don’t know enough to make much progress on a statement about the real efficacy of private market flood insurance. More study would be good. We can’t afford pig in a poke reforms.

But I do believe we can say some simple things. We wouldn’t have any reason to fear private insurance if we permitted it in a reasonable way. Some of the considerations:

• The affordability mechanisms of NFIP need to be sustained. For example, we need to sustain the HFIAA affordability surcharges and impose them on private market premiums. The surcharges turn out to have been an innovation in striking a balance between affordability and fiscal soundness in the NFIP program. We shall have reason to revisit these again in the future as we seek to keep a balance between affordability and fiscal soundness. But, whatever scheme emerges to sustain this balance between affordability and soundness in the NFIP should be imposed on private market flood insurance. Some level of surcharges on private market insurance surcharges contributed to the NFIP insurance fund would answer that need. This would prevent erosion of fiscal soundness in the NFIP fund when policyholders leave the government NFIP program to take up private market insurance. This is the most critical need to make private insurance fiscally safe and socially acceptable.
•  Congress should consider an additional levy on private premiums. This levy would be a contribution to the mapping and other data acquisition costs of FEMA. 
•  Finally, private insurance carriers should be required to share all information and data with the commons of both the NFIP pool and private insurance pool. This sharing of information would include any private mapping and modeling information that would help the entire market function better. This requirement would be in recognition that the private insurance carriers will be benefitting from the flood insurance mandate. The value of private sector investment in flood data shared with the flood insurance commons could be made creditable against the levy for contributions to maintaining flood maps and other data at FEMA.

We can make the market a part of our flood insurance future with modest safeguards. If we can make these new policies for fiscally safe and equitable private flood insurance, we ought to move ahead and bring private insurance into the service of those who live and conduct business in the flood plain. We can make private market insurance safe for the historic social goals of the NFIP. 

Technical Footnote

[i] For people who are buying NFIP, premium reductions are a good first order approximation of the benefits of the premium change. But, people who are not buying the insurance obviously have a lower value for it than the FEMA premium (or under postulate of rationality, they would be buying it now). Computing a premium reduction as a first order approximation for benefits for the people not buying the NFIP insurance is simply wrong. They would enjoy a second order benefit if being able to enjoy the computed new rate made them take up flood insurance. This is well known from economics. The standard consumer surplus argument is depicted in the diagram here. 

So subtracting the actuarial flood model premium from the NFIP premium overstates the value of the market insurance. The overstatement could be large as in the case where with the lower premium the homeowner still would not buy flood insurance. Then his benefits for private insurance would be zero.