Smoke Turns Distance Into an Economic Exposure
This chapter explains how wildfire smoke becomes a measurable household economic exposure outside the burn perimeter. The key task is not to repeat that smoke travels; it is to understand how the Philadelphia Fed study separates smoke from direct property destruction, what its heavy- and light-pollution comparison means, and why its estimates apply to a defined set of fires and distances rather than to every household under every plume.
A plume does not respect the perimeter
A fire perimeter answers a physical question: where did the ground burn? A smoke plume answers a different one: where did airborne particles travel, and at what concentration? Those boundaries can overlap, but they need not. Wind, topography, the duration of a fire, and the height and movement of emissions can carry smoke over communities whose structures remain intact. For a household outside the perimeter, the relevant exposure may be polluted air rather than a destroyed roof, but that exposure can still affect health, work, spending, and the ability to repay debt.
That distinction is why the Philadelphia Fed paper does not estimate its main smoke effects by simply comparing everyone near the Camp Fire with everyone farther away. People inside or immediately next to the burn zone may have lost property, evacuated, lost access to local services, or faced direct fire damage at the same time that they breathed smoke. If those households were included in a smoke estimate, a later increase in delinquency could not be cleanly attributed to particulate pollution. The study therefore removes the burn area and its immediate surroundings from the main distant-smoke comparison.
The primary comparison covers zip codes five to 30 miles from the fire boundary. Within that outer band, the researchers classify locations by the pollution they experienced after the fire. The heavier-pollution group generally consists of zip codes with average daily PM2.5 above 40 micrograms per cubic meter; the light-pollution group consists of zip codes below 10. PM2.5 is fine particulate matter small enough to be a health concern, but the important point for the financial analysis is that the study uses it to distinguish levels of wildfire-attributable exposure among households that are not themselves inside the burn footprint.
The middle of the pollution distribution is not the main contrast. By comparing relatively heavy and relatively light exposure, the study makes the treatment difference more visible than it would be if all locations were placed on one indistinct continuum. That choice has a trade-off. It clarifies whether a large smoke event is associated with changed outcomes, but it does not by itself describe the effect of every intermediate concentration. The estimate is tied to the contrast the researchers actually construct, not to an assumption that one threshold applies equally to every person or every day.
These are comparison groups, not labels for permanently polluted or clean communities. The study looks at outcomes before and after a wildfire event and asks whether the change for households in the heavier-pollution locations differs from the change for households in the lighter-pollution locations. That design makes the exposure contrast more concrete. A household in the heavy group is not treated merely because it lives in California or because it is within an arbitrary radius; it is treated because the event produced a defined pollution difference in its location relative to the comparison group.
The distance band also helps preserve the subject of the analysis. The question is not whether people who live farther from a fire are generally healthier or wealthier than people who live close to one. It is whether a wildfire-related pollution event changes financial outcomes among households in places that are close enough to experience substantial smoke but outside the area of direct structural destruction. The answer can then be compared with the direct-burn evidence without pretending the two treatments are identical.
The paper's main case is the 2018 Camp Fire, the most destructive event in its 2016–2020 sample. It also includes the Carr, Thomas, and LNU Complex fires in California. Concentrating on extreme events makes the exposure large enough to estimate with the available data, but it also creates an important boundary: the design tells us about households and fire conditions represented in those cases. It does not automatically describe the effect of a brief, light smoke day, a different fire regime, or a community with a different labor market and insurance system.
The exposure is also not a single measurement taken once. The study combines information about smoke, ground-level pollution, and wildfire-attributable particulate matter. That matters because an ordinary monitor reading can reflect sources other than a wildfire, while a modeled wildfire-attributable measure can capture the part of pollution the researchers want to assign to the fire. The combination is an attempt to connect a physical event to a household-level exposure with enough geographic and time detail to compare financial outcomes. It is not a claim that the data can observe every particle inhaled by every person.
What the comparison can and cannot isolate
The study uses a difference-in-differences framework. The phrase sounds technical, but its logic is straightforward. First, measure outcomes for the heavy- and light-pollution groups before the fire. Then measure how each group's outcomes change after the fire. The smoke estimate is based on the difference between those two changes. If credit delinquency rises more in the heavily polluted locations after the event than in the lightly polluted locations, that excess change is the candidate effect of the pollution exposure under the study's assumptions.
This logic is different from a simple before-and-after comparison. Credit delinquency could rise after a wildfire for reasons unrelated to smoke, such as a regional economic slowdown. It could also be higher in the heavy-pollution zip codes before the fire because those places differ in income or credit access. The design tries to account for both problems by comparing changes across groups rather than treating any post-fire movement as a wildfire effect. It also asks whether the groups followed similar patterns before the event, because a large pre-existing divergence would weaken the interpretation of a post-event difference.
The before-and-after structure matters because the two groups can differ in many ordinary ways at one point in time. One location may have a different baseline delinquency rate, age distribution, or credit mix. The analysis uses consumer or account fixed effects and time effects, along with time-varying borrower characteristics in relevant specifications, to focus more closely on changes for the same observed consumers or accounts over time. It also uses matching and additional geographic controls in some analyses. These choices do not create a perfect counterfactual, but they reduce the chance that a stable difference between places is mistaken for a wildfire effect.
A central problem is that measured air pollution can have more than one source. A change in PM2.5 after a fire could reflect wildfire smoke, but it could also be related to ordinary economic activity, weather, or another local condition that affects household finances. The authors therefore use more than one way to attribute pollution to the wildfire. One approach compares pollution in the fire month with pollution in the same month during the previous three years. Another uses wildfire-attributable PM2.5 estimates produced by the Stanford ECHO Lab. The paper also uses heavy smoke days as an instrument in its research design.
An instrument is useful only if its connection to the outcome is understood. Here, heavy smoke days help predict the rise in wildfire-related pollution, but the interpretation still depends on the assumption that they affect credit and labor outcomes through the smoke exposure rather than through an independent pathway. The study's use of the instrument is therefore a way to address a plausible source of endogeneity, not a permission to stop reasoning about assumptions. The alternative measures and fixed effects work in the same direction: they strengthen the smoke interpretation when results remain similar, but they cannot turn an observational setting into a randomized trial.
The purpose of these alternatives is not to produce several unrelated stories. It is to test whether the basic result survives when the exposure is defined differently. If estimates based on raw pollution changes and estimates based on wildfire-attributed pollution point in a similar direction, confidence in the smoke interpretation is stronger than it would be from one measure alone. Still, robustness is not the same as certainty. An instrumental-variable estimate depends on the assumption that the instrument changes financial outcomes through the relevant pollution channel rather than through some other route. The paper's design addresses important confounding concerns, but it is not a randomized exposure.
The outer-distance analysis provides another test of the perimeter model. The authors examine locations 30 to 100 miles from the Camp Fire and find smoke-related credit effects there as well, including higher credit-card and personal-loan delinquency in the reported estimates. They also analyze a much wider area of Northern and Central California affected by Camp Fire smoke. These exercises ask whether the result disappears as soon as a household is beyond the narrow 5–30-mile band. It does not, although the estimated effect is generally smaller at greater distance.
That distance pattern should not be interpreted as a universal law in which every extra mile reduces harm at a fixed rate. It is evidence that the economic footprint can extend beyond the place where direct destruction is measured. The amount of pollution, the duration of exposure, wind conditions, household characteristics, and the timing of work and health disruptions all matter. A 30–100-mile estimate is a result for the study's selected event and comparison, not a formula that can be applied to any map without new evidence.
The diagram captures a useful division of labor between geography and measurement. The perimeter identifies what the main smoke comparison removes. The plume identifies the wider exposure. The pollution groups turn that exposure into a contrast that can be analyzed, while alternative attribution measures test whether the contrast is plausibly tied to wildfire smoke. The outcome is not a map of every affected person; it is an estimate for households observed under a specified design.
That qualification is especially important when the result is translated into plain language. “Smoke affects household finances” is a supported summary of the paper's findings, provided it refers to the estimated relationship in the studied settings. “Every household downwind loses money” is not. Some households may have better ventilation, more flexible work, stronger insurance, more savings, or more credit capacity. Others may face health conditions or work arrangements that make the same pollution exposure more disruptive. The design estimates an average contrast across a population; it does not erase variation within that average.
The distance result also creates a scale distinction that will matter later. A direct-burn effect can be larger for each treated household because it includes property destruction, while a smoke effect can be smaller per household but reach a much larger population. A study estimate does not become unimportant because its individual coefficient is smaller. Nor does a large extrapolated total prove that every household experienced the average effect. The relevant questions are how many people were exposed, for how long, at what intensity, and how similar their circumstances were to the study population.
The study is also preliminary working-paper research, and the authors concentrate on four extreme California fires during a pre-COVID period. That focus is a strength for identifying a substantial event with detailed data, but it limits external validity. The economic consequences of smoke in an eastern metropolitan area, a rural community, or a place with different insurance and employment conditions could differ. The paper itself calls for further work on severe fires and smoke events in other regions. A careful reader should treat the estimates as evidence about a defined exposure pathway, not as a final national tariff on wildfire smoke.
The design nevertheless changes the central question in a productive way. Once the burn zone is removed, the household does not need to have lost a home for the wildfire to appear in economic data. The relevant mechanism may be a respiratory emergency, the cost of an air purifier, a missed shift, lower earnings, or an account payment that is delayed. The next step is to follow those pressures in the credit record: what changes first, how spending and repayment move, and whether the resulting balance and delinquency effects persist after the smoke begins to clear.