What the Evidence Actually Shows

The subject of this series is the post-flood reorganization of U.S. firms and plants: who exits, who rebuilds, who acquires released capital, and how financing shapes the result. The final question is why productivity can rise after a disaster without making the disaster beneficial. The evidence supports a conditional answer and a practical judgment: destruction can trigger selective reallocation, but only a chain of rebuilding, capable entry, and accessible finance can turn that disruption into a productivity improvement—and even then, the improvement is not the same thing as broad welfare.

The answer is easiest to lose when the story is compressed into one headline. “Productivity rose after the flood” hides the population being measured, the plants that disappeared, the equipment that changed hands, the firms that could not obtain credit, and the households whose losses are not in the productivity statistic. The evidence becomes clearer when those elements are kept in their causal order.

The complete mechanism is conditional

A flood first creates harm. Machinery can be damaged, inventories lost, production interrupted, and viable businesses forced to close. The working paper studies large, infrequent U.S. floods and finds that the probability of a manufacturing plant exiting within two years rises by 3.6 percentage points after a declared flood. The increase is not evenly distributed: plants in the bottom quartile of pre-flood TFP face an additional 1.1 percentage-point exit likelihood relative to plants in the upper three quartiles.

That is the destructive margin. It matters because an average calculated after the event may contain fewer low-performing plants than the average calculated before it. A later increase in the productivity of operating plants can therefore be partly a composition effect. But calling the exit pattern “cleansing” does not mean that every plant that closed was wasteful, or that the loss of jobs, output, and community capacity was harmless. A flood can remove a low-performing plant and destroy a high-performing one in the same region. Selection describes the relative pattern, not the total human cost.

The second part of the mechanism occurs among plants that remain. Survivors show persistent increases in capital retirement, along with repair and refuse-removal activity. The evidence is consistent with flood-damaged equipment being disposed of or sold rather than simply aging according to an ordinary depreciation schedule. Ferrous-metal processors provide a physical check: because moisture accelerates rust, these plants show greater flood-related inventory losses and subsequent employment and payroll reductions relative to processors of more corrosion-resistant metals.

Retirement is not the end of the survivor story. Normalized net investment rises by about 1.5 percent after declared floods, and county-level investment also increases. That means surviving plants are not merely writing off damaged assets; on average, they invest enough to more than offset the capital they retire. Their TFP then rises persistently, reaching more than 11 percent above the pre-flood baseline by the sixth post-flood year.

The right interpretation combines selection and upgrading. Some of the survivor gain exists because weaker plants are more likely to exit. Some may come from replacing damaged equipment, reorganizing production, or adopting a better process. The evidence does not allocate the 11-percent increase between those channels. Nor does it say that every surviving plant modernizes. Survivor estimates are conditional on remaining in business for at least three additional years, and capital measures can be affected by imputation in the underlying manufacturing survey.

The creative margin begins when capital and opportunity move beyond the original incumbent. An exiting plant may release a usable machine, while a survivor may sell equipment it replaces. A young entrant may acquire that used capital because it cannot afford a new machine. The post-flood single-unit entrant evidence fits this sequence: the share of machinery investment devoted to brand-new machines falls by 1.9 percentage points, while entrant TFP is about 14 percent higher than among pre-flood entrants.

This combination is important precisely because it is not a simple story of replacement with the newest technology. Used capital can be the affordable way for a productive entrant to begin operating. Yet the result is not proof that every machine came from a closing plant, that every entrant used it well, or that used equipment caused the TFP difference. Entrants are observed only once in this analysis, and the new-versus-used breakdown relies on earlier survey waves. The evidence supports a plausible and measured resource-transfer pattern, not a complete transaction history.

Financing determines whether that pattern can scale.