How Much Is the Recovery Worth?
The subject here is the distance between firm-level reallocation and community-level recovery: whether changes in plants’ use of capital after a flood appear in aggregate productivity, income, and public finances. The specific task is to read those outcomes as different signals and then reconstruct the paper’s marginal value of public funds calculation. That approach supports a conditional judgment about disaster assistance—one that counts fiscal returns without pretending that a positive aggregate estimate erases losses or proves that rebuilding is always the right choice.
The distinction matters because a policy can succeed on one dimension and fail on another. Capital may move toward more productive plants while some households lose income. County income may rise while the average plant-level TFP estimate remains imprecise. Tax revenue may exceed modeled public cost while the calculation says little about distribution, safety, or whether residents should remain in a repeatedly exposed place. The paper’s aggregate and fiscal evidence is most useful when each result is allowed to answer its own question.
Aggregate productivity is not a single score
The aggregate analysis begins with plants operating in the same county and narrowly defined industry. That comparison keeps the question focused: after a declared flood, does capital appear to be allocated more evenly across similar production units? The paper uses several measures because no single statistic can distinguish a better allocation from a change in the composition of plants.
Mean TFP is the most intuitive measure. It asks how productive the average operating plant is, given the inputs measured in the data. After declared floods, the mean-TFP estimate is close to zero or slightly positive and imprecisely estimated. That is not a failure of the analysis. It tells us that the paper does not establish a large, precisely measured increase in the average productivity level for the aggregate comparison cells. A stable mean can coexist with major plant turnover and with improvements in how capital is distributed.
TFP dispersion adds another perspective by asking how far apart plants are in measured productivity. The paper reports a directionally lower spread after declared floods. A narrower spread can be consistent with productive reallocation, but it is not conclusive on its own. If the most productive plants were damaged or stopped experimenting, the top of the distribution could fall toward the middle. If the least productive plants left, the bottom could rise. The meaning of a smaller spread depends on the exit and entry evidence that identifies which tail is changing.
The paper’s more direct allocation indicator is the interquartile range of output relative to capital stock, reported as MRPK dispersion. The intuition is that similar plants with very different output per unit of capital may be using capital unevenly. In a frictionless allocation, capital would tend to move toward firms that can generate more from it until the marginal returns were more nearly equal. A narrower MRPK distribution is therefore consistent with capital moving toward better uses.
The magnitude is modest but informative. The baseline two-way fixed-effects estimate shows about a 1.1 percent decline in the interquartile gap. The Sun–Abraham estimate, which addresses concerns about staggered event timing, shows a larger decline of about 2.7 percent. The difference between the estimates is itself useful evidence about uncertainty.