The Unreturned

More than $110 billion in federal resources — $118 billion counting tax relief. Over $77 billion already dispensed or made available. Nearly $6 billion to the Army Corps of Engineers for levees, 220 miles of floodwall repaired. 103 of 122 million cubic yards of debris removed. $16.7 billion through HUD block grants, described as the largest housing recovery program in American history. $10.4 billion in Small Business Administration disaster loans. Twelve presidential visits; thirteen by the First Lady. Emergency stockpiles increased 300 percent for meals, 400 percent for ice.

Every figure is an input. Dollars appropriated, cubic yards hauled, miles repaired, visits made. Not one is an outcome, and not one is a person. The document measures governmental exertion — the single variable the government fully controls — at a moment when the outcome variable was already known and pointing the other way. New Orleans had roughly 455,000 residents before the storm. At the anniversary it had roughly 230,000.

That a fact sheet flattered its author is not the interesting part. That is what a fact sheet is for. The interesting part is that the largest housing recovery program in American history had a formula inside it, and the formula was doing the actual work.

Louisiana's Road Home program paid homeowners the lesser of the cost to repair the house or its pre-storm market value. Read that twice; it is the entire story. A house's pre-storm market value in New Orleans in 2005 was a number produced by seventy years of redlining, appraisal convention, mortgage access, and the ordinary compounding of all three. Indexing a rebuilding grant to that number does not require anyone to intend anything. It only requires the arithmetic to run.

It ran. One plaintiff in the litigation that followed received $1,400 to rebuild a house with an estimated $150,000 in damage. The Greater New Orleans Fair Housing Action Center and the National Fair Housing Alliance sued in 2008 on behalf of a class potentially exceeding 20,000 homeowners. HUD settled in 2011: $473 million in supplemental awards, recalculated on damage cost rather than market value.

The settlement conceded the point. It conceded it six years after the storm — which is to say, several years after every family who could not wait had already signed a lease elsewhere, enrolled their children elsewhere, and stopped being a New Orleans household in any sense a census can detect. A remedy that arrives after the migration is not a remedy. It is an accounting correction applied to people who have already been sorted.

And Road Home only ever spoke to owners. Most displaced Black New Orleanians were renters, and the program had no sentence for them — a grant indexed to the value of an asset has nothing to say to a person who holds none. It did not discriminate against renters. It did not reach them. That is the deeper version of the same finding: the fact sheet counted no people, and the largest housing recovery program in American history counted only owners. Having nothing to say about someone is a quieter way of deciding they will not be back, and it produces no plaintiff class to say so.

The renters' side of the ledger did have authors, which is worth conceding plainly. The demolition of the Big Four public housing developments in 2007 and 2008 — thousands of units, much of the stock structurally sound, replaced by mixed-income projects holding a fraction of the deeply subsidized apartments — was decided in public, over organized objection, by people who defended it at the time and afterward. So was the wholesale conversion of the school system. Those were arguments, and they were won. The formula was never argued over, because there was nothing in it to argue with. Two tracks, one destination. The second is the one worth writing about precisely because nobody had to win it.

Two decades on, the sorting is visible in the count. New Orleans had roughly 325,000 Black residents in 2000 and roughly 204,000 in 2024 — a loss exceeding 120,000. White residents went from about 136,000 to about 113,000, a loss of about 23,000. Thirty-seven percent against seventeen. The city that came back is smaller, whiter, and wealthier, and none of those three facts is an act of God. The storm was the event. The formula was the boundary condition.

Notice that the fact sheet and the formula are the same instrument appearing twice in the same recovery. The document substituted an administratively legible number — dollars, cubic yards, visits — for the thing that mattered. The grant substituted an administratively legible number — market value — for the thing that mattered. A state that can see only what it already counts will rebuild only what was already counted, and will then produce, in complete good faith, a ledger showing the operation succeeded.

So the structure worth naming, because it is still in service: disaster relief indexed to pre-existing value reproduces pre-existing value. That is not a failure of the program. It is the program executing correctly, which is a considerably worse problem — failures get fixed, correct operation gets replicated. Insurance logic restores the insured to their prior position, and it is neutral only if the prior position was. The neutrality of the formula is precisely what made it efficient at converting temporary displacement into permanent rearrangement.

Anniversaries are attended by the people who could come. That is a tautology, and it is also why a commemoration is structurally incapable of counting what it commemorates. In August 2006 a president stood in a city and marked what had happened to it. The 225,000 people who were not there held a more accurate view of the recovery than anyone on the podium, and no mechanism existed by which their view could enter the record.

The pattern is still running, so: the next major American disaster recovery will again index household assistance to pre-loss asset value, it will again be defended as neutral on the grounds that a formula cannot hold a motive, and the demographic consequence will again be published about a decade later by a research center, after the population it describes has finished not returning.