Why Destruction Alone Does Not Raise Productivity
The subject here is the post-flood productivity of firms and plants, not the abstract idea of economic “resilience.” The specific question is what must happen after low-performing plants are pushed out for a disaster to be followed by better resource allocation. The answer gives us a disciplined way to read the evidence: exit can change who is counted, but only reallocation can explain why the resources of the plants that leave are used more effectively.
That distinction is the difference between the two meanings hidden in “creative destruction.” The destruction is observable when a plant closes, machinery is scrapped, inventory is lost, or production stops. The creative part is a further claim about where capital, workers, and opportunities go. It requires a recipient: a surviving plant that upgrades, or a new firm that acquires the equipment and can put it to productive use. Financing determines whether that recipient can act.
“Cleansing” describes selection, not benefit
In ordinary economic discussion, “cleansing” can sound like a positive judgment. In the flood research, it is more precise and more limited. A negative shock may remove firms that were already less productive than their industry peers. If those firms leave while higher-productivity firms remain, the average productivity of the operating population can improve. The term describes the selection pattern. It does not say that the shock was desirable, that the plants were disposable, or that the people affected by their closure are better off.
The working paper’s exit result illustrates the distinction. After a declared flood, a plant’s probability of exiting within two years rises by 3.6 percentage points relative to the pre-flood period. Among plants observed before the flood, those in the bottom quartile of total factor productivity—TFP—face an additional 1.1 percentage-point increase relative to plants in the top three quartiles. The comparison is meaningful because it is about the productivity ranking of plants before the event, not a ranking reconstructed after some plants have already disappeared.
TFP is useful here because it summarizes how much output a plant produces from its measured inputs, such as labor and capital, after accounting for the fact that plants combine those inputs differently. But it is not a moral score, and it is not a complete balance sheet. A high-TFP plant can still be vulnerable to water damage. A low-TFP plant may provide important jobs or services. If a productive factory closes because a flood destroys its machinery, the resulting selection may raise the average of survivors while reducing the community’s productive capacity in a way that matters to households.
The same result can therefore support two statements that sound contradictory but are not. First, the flood selectively removes more low-performing plants than high-performing plants. Second, the flood harms businesses and workers, including some productive ones. The first is a statement about the composition of production. The second is a statement about the shock’s costs. Confusing them is how a statistical description becomes an unwarranted celebration of destruction.
Three measures, three different questions
A reader also needs to separate the measures used to describe the post-flood economy. The mean of plant-level TFP asks: how productive is the typical or average operating plant? It can move because individual plants improve, because low-performing plants exit, because high-performing plants enter, or because several of these changes occur together. A mean is a level indicator, but it does not reveal which mechanism produced the change.
TFP dispersion asks a different question: how far apart are plants in productivity? A narrower distribution can be consistent with improved allocation, but it can also result from a loss of highly productive plants or from a collapse of experimentation at the top. A wider distribution can reflect inefficiency, but it might also accompany productive experimentation. Dispersion is informative only when read alongside the direction of the mean and the identities of the plants entering and leaving.
The paper also examines dispersion in output per capital stock, reported as an MRPK measure. The intuition is straightforward. If two similar plants in the same county and industry have very different output relative to their capital, capital may not be in its best uses. In a frictionless benchmark, the marginal productivity of an input would be more nearly equal across firms because capital would flow toward the firms able to generate more from it. A smaller gap in this measure is therefore evidence consistent with better allocation of capital.
That is still an indirect indicator, not a direct welfare measure. A narrower MRPK gap does not tell us whether the region lost jobs, whether income was distributed fairly, or whether people wanted to remain there. Nor does it prove that every machine moved to a better owner. It says that the observed distribution of capital productivity became less uneven within the comparison cells. To interpret that as productive reallocation, we need plant-level evidence showing who released capital and who received it.
The difference can be seen with a simple accounting example. Suppose ten plants produce with very different levels of efficiency. A flood closes the two weakest plants, but it also destroys one of the strongest. The mean TFP among the eight survivors might rise or fall depending on the magnitudes. The spread might narrow because the extremes disappeared. Neither outcome tells us whether the machines from the closed plants were sold to capable entrants, left idle, or lost entirely. The identities and subsequent investment decisions matter more than the average alone.
Exit is necessary but not sufficient
Selective exit can be one part of an efficiency-enhancing process. It removes a producer whose pre-flood performance suggests that the same resources might generate more elsewhere. Yet exit does not transfer resources by itself. A closed plant’s machinery may be damaged beyond use. Its usable equipment may be sold at a discount, but potential buyers may lack cash. Its workers may leave the area. Its building may sit empty. The creative margin begins only when another producer can take up some of what has been released.
This is why a complete sequence has at least four links. A negative shock first changes the survival decisions of incumbent plants. Some low-productivity plants exit at higher rates. Their capital, space, equipment, and workers then become available—or are destroyed. Survivors and entrants decide whether to repair, replace, buy, or abandon those resources. Finally, the productivity of the new arrangement determines whether the region has experienced reallocation or merely contraction.
Each link can fail. Exit may not be selective: a flood can close firms according to damage exposure or liquidity rather than pre-flood productivity. Resources may be unusable, stranded, or purchased by firms that cannot operate them well. Survivors may replace equipment with technology that is no better suited to production. Entrants may be less productive than the firms they replace. A post-flood average can still look different, but the chain would not justify calling the change an improvement in allocation.
The distinction is especially important for used capital. A machine that is second-hand is not necessarily obsolete. It may be exactly the affordable input a young firm needs to begin production. Conversely, a cheap machine may be a poor investment if it is damaged, incompatible with the entrant’s process, or too costly to maintain. The relevant evidence must therefore connect the source of the equipment to the productivity and survival of its new user, rather than treating the price or age of the machine as sufficient proof.
Survivors supply another part of the creative margin. A plant that remains open may retire damaged machinery and install new equipment. Its later productivity may improve because the plant was already stronger, because it upgraded, or because both are true. That is why survivor productivity cannot be interpreted without knowing the selection rule: an average among plants that remain for several years is not the average effect on all plants present before the flood.
Why finance changes the quality of churn
Financing is the link that makes the sequence more or less likely to complete. Flood recovery creates an unusual financing problem. Firms need to spend before normal revenue returns, while collateral may be damaged and the future demand for their output may be uncertain. A young entrant faces an even sharper constraint: it has little operating history, few assets, and no established cash flow with which to fund a new machine. If credit is unavailable, the firms most capable of using released resources may be unable to enter or rebuild.
This means finance can change the quality of churn, not merely its quantity. With adequate liquidity, a flood may produce more exits and more entry at the same time: weak plants leave, while capable survivors and entrants obtain capital. Without adequate liquidity, the same shock may produce closures without productive replacement. The economy experiences the destructive margin but not the creative one. Credit can thus affect selection on both sides of the market—who can survive and who can become a new producer.
The contrast in the working paper supports this logic. In severe non-declared flood settings, exits also rise, but the additional exit probability of low-productivity plants is not statistically distinguishable from that of higher-productivity plants. New entrants show no positive productivity selection, and mean TFP falls by about 0.73 percent in the reported baseline estimate. The paper interprets this combination as a sign that financing constraints can suppress experimentation and productive turnover. The result is not that no firm can recover without federal assistance. It is that the pattern of recovery lacks the selective entry and exit observed in the declared-flood setting.
Public finance introduces a trade-off rather than a free improvement. A low-interest loan can keep a viable plant from liquidating and let a young firm buy equipment. But if aid merely preserves an inefficient incumbent, it can prevent resources from moving to a better user. The question is therefore not whether government spending is large, or whether lending rises in aggregate. It is whether the firms receiving or benefiting from the added liquidity are able to convert it into production, investment, and jobs more effectively than the alternatives left without finance.
This is also why the paper’s evidence must be read as a mechanism rather than a slogan. The observed loan and investment patterns can be consistent with financing relaxing a constraint, but they do not mean that every additional dollar causes a productive entry. The non-declared comparison is informative but not a randomized switch for federal aid; the events can differ in composition even after efforts to make their severity more comparable. Strong inference comes from the convergence of several observations—selective exits, capital adjustment, entrant outcomes, lending patterns, and aggregate measures—not from any one coefficient.
A causal map for reading the result
The logic can be summarized as a sequence with a genuine failure path:
The diagram does not say that financing guarantees efficient recovery. It shows where the mechanism can break. A flood may produce selective exit, but if resources are destroyed or finance is constrained, the sequence ends before productive reallocation. If survivors and entrants can obtain capital and operate it effectively, the mean may recover and the dispersion of productivity or capital returns may narrow. Those are different outcomes, and both remain distinct from the welfare question.
The next question is therefore concrete: what happens inside an actual plant after the water recedes? To establish the creative margin, we need to see whether survivors retire and replace damaged capital, whether their investment exceeds ordinary adjustment, and how their later productivity changes. Only then can the abstract sequence be tied to the physical decisions that make recovery possible—or leave it incomplete.