Impacts – 09 Study Limitations and Error

Impacts of Visitor Spending on Local Economy:
Crater Lake National Park, 2001
Study Limitations and Error


The accuracy of the MGM2 estimates rests on the three inputs: visits, spending averages, and multipliers. The MGM2 generic “rural area” multipliers were selected to best represent the economic character of the region. Visitor segment shares and spending averages are derived from the 2001 Crater Lake National Park Visitor Survey. These are subject to sampling errors, measurement errors and some seasonal biases as we had to project to annual totals from the summer season VSP study.

The Crater Lake Visitor Survey was conducted during a 7-day period at selected locations during August, 2001. Several adjustments were made to the VSP survey results to correct for likely sampling biases and to better represent year-round visitation. First, cases were weighted inversely to the park re-entry rate to correct for greater probabilities of selecting multientry visitors. Second, the segment shares for overnight visitors staying at hotels and campsites inside the park were adjusted to be consistent with park overnight stay data in the Public Use Statistics. Third, spending averages were adjusted downward to reflect lower off-season rates9. Length of stay, party size and re-entry rates were assumed to not vary by season. Finally, the shares for visitors staying at campsites outside the park were reduced by 20% to reflect less camping during the off-season.

The sampling errors for the estimates of the average spending per party night were 5% overall and ranged from 5- 18% for individual segments10. Spending averages can also vary by about 10% based on decisions to treat missing spending data as zeros or not, and how many and which outliers to delete11. Our analysis omitted cases with missing values in all spending categories and also cases with spending of more than $1000 per party per day.

Depending on the direction and magnitude of errors in visits, spending, and multipliers, the errors may compound or cancel each other. The most important potential errors are in the estimates of visits and segment shares. As the model is linear, doubling visitors will double spending and impacts. Errors in other parameters, such as re-entry rates, lengths of stay and party sizes, also directly translate into errors in party nights, which are multiplied by the spending averages. Using a 95% confidence interval on the spending averages and total recreation visits in 2001, the park visitor spending is estimated to range between $27.6 million and $33.7 million in 2001.

In addition to these statistical issues, there are also conceptual issues regarding how much and which spending the park may claim. Around 75% of park visitors indicated that Crater Lake NP was their primary destination (Visitor Service Project, 2001). Some spending on trips where Crater Lake NP is not the primary destination, such as visiting friends and relatives or staying at seasonal homes, likely would not be lost to the region if Crater Lake NP were closed. Only one day’s worth of spending is attributed to the park for visitors whose primary trip purpose was visiting friends/relatives or staying at seasonal homes.

Local visitors are often excluded in estimating economic impacts, but have been included here. Since they are a distinct segment, their contribution to the totals is readily estimated and subtracted from totals, as desired. Locals accounted for about $1 million or 3% of overall visitor spending. If these visitors would have gone outside the region in the absence of the park, the spending would be lost to the region.

<< previousnext >>