Abstract
Glasgow and Train propose a correction to the government’s estimate of recreation damages in the Deepwater Horizon oil spill assessment. Their correction depends on a description of behavior they assume to be true but did not test. We find that evidence on recreators’ response to the spill contradicts their assumption and that the standard welfare calculations applied in the case were appropriate. (JEL D61, D81)
1. Introduction
Glasgow and Train (2018) examine the value of effects on recreation from an incident that has the potential to cause environmental harm at a particular site but has not. People often avoid sites in these circumstances, and the issue has traditionally been viewed as a matter of perception: the quality of sites in the area of an incident is perceived to be degraded even if no harm has occurred at some sites (Hausman, Leonard, and McFadden 1995; Byrd et al. 2001; Athos Delaware River Trustee Council 2007). In this traditional view, the perception of a quality decline may lead to smaller effects on demand for recreation than actual environmental harm, but the welfare calculations following from the change in demand are the same.
Glasgow and Train (hereafter GT) argue instead that people in this situation are responding to imperfect foreknowledge about site conditions, a problem that is resolved upon visiting a site. They appear to have in mind two possible cases, one where imperfect foreknowledge arises from advance planning of trips when future conditions are unknown, the other where people have uncertainty or false impressions about current conditions and need to see for themselves. In either case they conclude that those who avoid an unharmed site suffer a loss, but those who visit the site do not. They further argue that standard welfare measures overstate losses in these circumstances by incorrectly ascribing a loss to people who visit the site. As an example they point to the Deepwater Horizon oil spill in the Gulf of Mexico, in which oil spread over hundreds of miles of Gulf Coast shoreline but did not reach the Florida Peninsula. The authors propose a correction to standard welfare measures and apply it to the government’s estimate of welfare loss in the Florida Peninsula, an estimate that we and other experts developed on behalf of federal and state agencies during the Deepwater Horizon oil spill assessment (English and McConnell 2015).
We agree with GT that their corrected welfare measure could in principle apply in some cases. However, we conclude that any such adjustment would be small in the Deepwater Horizon case. Below we elaborate on the following points:
GT’s description of behavior, whereby potential Gulf visitors avoided unoiled sites due solely to imperfect foreknowledge about site conditions, is not a definitive conclusion from economic theory. Rather, it is a hypothesis about preferences and behavior that needs to be tested. An alternative hypothesis is that people were responding to a perceived decline in quality, such as a belief the water might be contaminated. The extent to which people were responding to imperfect foreknowledge rather than perception is the extent to which any welfare correction may be warranted.
People living close to shore in the Florida Peninsula were unlikely to face the problem of imperfect foreknowledge, since they could take trips without advance planning and could personally view beach conditions with little cost or effort. However, our data show a significant number of lost trips by those living close to Peninsula beaches. This indicates that many people were responding to a perceived decline in quality rather than imperfect foreknowledge about beach conditions. While GT apply their behavioral model and the resulting welfare adjustment to all people, evidently it applies at most to some people.
If the GT hypothesis applies to any significant number of people, then it should affect the geographic pattern of losses observed at unoiled sites relative to oiled sites. Expanding on point number 2, behavior based on imperfect foreknowledge should lead to fewer canceled trips close to shore for unoiled sites, and more canceled trips as distance and the attendant problem of imperfect foreknowledge increase. We looked at the relationship between distance and recreation impacts in the Florida Peninsula and compared it to the relationship between distance and recreation impacts on the Gulf Coast west of the Peninsula, where sites were oiled soon after the spill began. The two appear to be quite similar. This indicates that the number of people responding to imperfect foreknowledge at unoiled sites was too small to have any detectable impact on the pattern of losses in the Peninsula. Any correction to welfare estimates for unoiled sites would therefore also be small.
Given the popularity of the Florida Peninsula as a recreation destination and the large number of visitors coming from afar and planning trips in advance, the Deepwater Horizon spill may be a good test case for the GT model. The absence of support in the Deepwater Horizon case for the alternative welfare measure proposed by GT is likely to imply similar conclusions in most oil spills. Perceptions rather than imperfect foreknowledge appear to drive behavior, and standard welfare measures are generally appropriate for all sites where a decline in recreation activity has occurred.
2. Perceptions, Imperfect Foreknowledge, and Implied Behavior
While the economic issue is potentially broader, for simplicity we focus on oil spills and the Deepwater Horizon case. It is possible that people avoided unoiled sites on the Florida Peninsula because they perceived the quality of sites to be diminished. This perception could encompass many possible factors. People may have had fears about water quality related to chemical dispersants that were widely used during the spill. People may have been uncertain about the quality or timeliness of available information about oiling or about where to find reliable information. This problem would not be resolved upon visiting a site if people lacked confidence in their own ability to evaluate whether oil was present at a beach; they might fear that oil could be mixed in the water or sand in ways that are not apparent to an inexperienced observer. Some people may have wanted to “play it safe” and were simply not comfortable using beaches on the Gulf of Mexico when a massive oil spill had recently occurred there.
In a telephone survey we conducted with residents living in states adjacent to the Gulf, many respondents said they canceled trips because of the spill, but they were not specifically asked why the spill deterred them. However, an open-ended question at the end of the survey asked respondents to describe in their own words how the spill affected their recreation. Numerous respondents expressed concerns about oil and dispersants in the water, including respondents living in the Florida Peninsula. Such concerns are consistent with research showing people often react with visceral disgust to substances that could be perceived as contaminated, particularly substances that come in contact with their mouths or skin (Rozin, Millman, and Nemeroff, 1986; Tybur et al. 2013). This type of reaction could be of particular concern to those who said they refrained from swimming in the Gulf following the spill.
Perceptions about environmental quality may have a smaller effect than actual environmental harm, but otherwise the resulting changes in behavior and value are the same. For those concerned about it, the value of trips declines and they go less often or not at all. The resulting inward shift in the demand function reflects the value of lost trips for those who do not go and the potentially diminished value of trips for those who do. In fact, any quality decline is fundamentally a matter of perception, as even visible oiling does not deter some beachgoers.
Apart from perceptions, people may also have avoided unoiled Peninsula beaches because of imperfect foreknowledge about site conditions. This is GT’s hypothesis, and if true it would imply the need for an adjustment to welfare estimates, as they suggest. However, it has a key implication for behavior that differs from the influence of perceptions. Specifically, the farther people live from the beach, the higher the expected effect on demand for trips. This is because at greater distances the time horizon for advance planning is usually longer and the risk of a change in future conditions is therefore greater. The cost of resolving uncertainty or misimpressions by observing conditions in person is also greater.
It is worth noting that the issue of unoiled sites arises not only in locations outside an area where oiling has occurred. Within an area of oiling, people often avoid all locations even if some of them are free from oil. Some people also continue to avoid areas that were oiled even after oil has been removed and any closures or advisories have been lifted.
The question of perceptions versus imperfect foreknowledge is the key issue in evaluating whether traditional welfare estimates are appropriate for unoiled sites. GT emphasize the question of whether people at unoiled sites suffer a loss, but this is a red herring. If people are responding to perception and some are concerned about using unoiled sites and others are not, then the majority of people who go to the beach may be from the second group and would not be affected. But there would be no reason to adjust welfare estimates in this case. Traditional welfare estimates encompass this situation and do not assume people onsite necessarily suffer a loss.
3. Lost Trips by People Close to Shore
As described earlier, if people are responding solely to imperfect foreknowledge about site conditions, then those living close to shore would be unlikely to take fewer trips to unoiled sites. GT appear to acknowledge this. The model they use to illustrate their proposed welfare adjustment includes only overnight trips that “could be expected to have been planned in advance” (Glasgow and Train 2018, 91). To ensure this, they examine impacts to Gulf Coast beaches using a population that excludes residents of eight southeastern states, many of whom live hundreds of miles from the Gulf Coast.
While we agree that any correction to welfare estimates should not be applied to people close to shore, this leaves unresolved the salient issue that these people often cancel a substantial number of trips. In the Deepwater Horizon assessment, residents of the eight-state area delineated by GT accounted for 2.8 million lost recreation days to the Peninsula, or 71.4% of all lost Peninsula days (the spill effects were measured onsite using recreation days rather than trips; there was an average of 1.7 recreation days per trip). For reasons that are not clarified in their article, GT introduce this eight-state exemption in their hypothetical scenarios but do not carry it over to their proposed correction of our Deepwater Horizon welfare estimates. If they had, our estimated welfare loss in the Florida Peninsula of $159 million would not be revised to $20 million, as GT calculate, but to $118.2 million instead (assuming a constant value for recreation days and applying GT’s formula to 1.1 million lost days and 6.2 million baseline days outside the eight states). The result is a decline of only 25.8% in the total Peninsula welfare loss.
Yet even this adjustment to GT’s calculations does not resolve the problem. The decline in the demand for trips in areas close to shore directly contradicts GT’s hypothesis about preferences and behavior. During the eight months when recreation was reduced in the Florida Peninsula, there was a 23.2% decline in recreation days by residents living within 10 miles of the Peninsula shoreline. Evidently a significant proportion of people were responding to their perceptions rather than to imperfect foreknowledge. This segment of the population is revealed directly for those living close to Peninsula beaches, where perception is the only plausible explanation for their response, but the same preferences exist throughout the country. (The consistency of preferences across space, conditional on demographic variables, is fundamental to the travel-cost model, and there would be no theoretical or empirical basis for challenging it in this instance.) This means that the revised correction of 25.8% would need to be further reduced: it would need to account for the fact that even among those who live far from shore, many would have canceled trips just on the basis of a perceived quality decline without the need to personally evaluate present or future conditions. If lost trips throughout the country are entirely accounted for by this segment of the population, then no welfare correction is warranted.
4. Comparing Losses at Oiled and Unoiled Sites
If most people are responding to imperfect foreknowledge, the rate of canceled trips should start near zero close to shore and increase with distance from the shore as imperfect foreknowledge becomes more of a problem. If any significant portion of people are responding to imperfect foreknowledge, then greater distances should at least be associated with a detectable increase in trip cancellations, or failing that, an increase relative to the response at oiled beaches. The comparison to oiled beaches may control for any confounding factors that could make the expected increase in Peninsula effects hard to detect if there are countervailing influences correlated with both distance and response to the spill.
Figure 1 shows the relationship between the distance recreators live from the Peninsula shore and the rate at which they curtailed their recreation at Peninsula beaches. The same relationship is also shown for areas of the Gulf where shoreline oiling occurred, a region we term the “North Gulf” and that includes beaches in northwestern Florida, Alabama, Mississippi, and Louisiana. As in the Deepwater Horizon assessment, lost recreation days were calculated by comparing beach visitation following the spill with visitation in the same locations after activity had returned to normal. The assessment relied on 500,000 aerial photographs, 35,000 onsite counts, and 129,000 onsite interviews collected between June 2010 and May 2013 (Tourangeau et al. 2017). To create Figure 1, the origin of trips was taken from the onsite interviews, and the distribution of trip origins across distance bands was applied to both ground and aerial counts in the spill and postspill periods, respectively. The results shown are for June 2010 through January 2011, the period when losses were found to occur in the Florida Peninsula.
Percentage Decline in Recreation Days by Recreators’ Distance from Shore
The decline in recreation days for residents within 10 miles of the Peninsula shore is not zero, but 23.2%, as noted earlier. This is slightly higher than the 22.2% decline in Peninsula recreation days from all origins throughout the country. Indeed Figure 1 illustrates that as distance from shore increases, the rate of loss is flat or declining over most distance ranges. Only in the range of 50 to 150 miles does the proportion of lost days increase. An increase in losses over such a limited range offers little support for the hypothesis of imperfect foreknowledge, and in any case a similar increase appears in the North Gulf over the range of 20 to 50 miles. Overall, impacts at the unoiled beaches of the Florida Peninsula are lower than impacts at the oiled beaches of the North Gulf, consistent with the expectation that a perceived quality decline has less of an impact than actual oiling. While a more detailed investigation could control for the relationship between distance and other factors that could affect the decline in trips, such as the availability of substitute sites or demographic characteristics, the analysis in Figure 1 seems sufficient to evaluate the issue at hand. We see little distinction between the spatial pattern of losses in the North Gulf and the Peninsula, and little indication that imperfect foreknowledge is a significant factor affecting demand or welfare at unoiled sites.
5. Conclusion
We find little evidence to support GT’s assumption that those who were concerned about sites where oil never came ashore simply needed to visit the sites to be assured that conditions were fine. If some people did fit their assumed description of behavior, it would require methods more sophisticated than what they propose to uncover this relatively small group and calculate their adjusted loss. We conclude that our methods and results in the Deepwater Horizon case were sound, and that conventional methods for welfare estimation are likely to be appropriate in future cases.
Acknowledgments
We thank Norman Meade for helpful comments and suggestions.
This article requires a subscription to view the full text. If you have a subscription you may use the login form below to view the article. Access to this article can also be purchased.
