Tuesday, January 30, 2007
Surveying the Tourists
At yesterday's SOVA symposium, I listened to a presentation regarding research. I responded with the following email:
I enjoyed learning about the Dennett Group's visitor research for Southern Oregon. I want to follow up on the brief conversation we had about potential problems.
First consider the following simple model. Cities A and B each have 1,000 population. Everyone in A visits Southern Oregon once a year. 100 people in B visit Southern Oregon 10 times a year and no one else. If you sample the populations of A and B at home, you will decide that 10X as many SO visitors are from A as B. If you conduct intercepts in SO, you will get the same number of A and B. Longwoods is based on the former, and you on the latter. You won't get the same results.
You remarked that the only problems to worry about would be systematic sources of error. Here are five.
1) Assume that there is a decreasing likelihood on any particular trip of someone going to a Visitor Center. Your method will understate the number of trips to SO that a particular visitor takes.
2) However, since the cumulative likelihood of being at a visitor center at least once increases with more trips, you will paradoxically overweight the demographics of people who visit SO repeatedly.
3) You will underweight low incomes, because such people are less likely to base decisions on reading matter, which is what they get in visitor centers.
4) You will underweight "visiting friends and relatives" because such people have a built-in source of visitor information that is not available to FIT folks. They are consequently less likely to enter a visitor center.
5) You will underweight people who get their information from the Internet. Someone who has printed a local map from Mapquest doesn't need to get one from the visitor center.
You have a very accurate survey underway of exactly what it appears to be -- the profile of the average person who walks into a visitor center, or to a small degree an attraction. You also have useful information for the lodging, shopping, and dining folks, but it certainly won't be definitive.
I have devised a strategy for doing an online survey for the Oregon Coast. It will have errors, as does yours, but it will give useful information. It will say less about visitors to visitor centers but more about occupants of hotel rooms. I will be curious to see what differences appear between your results, my results, and Longwoods.
Rob Spooner, Publisher
Oregon Coast Magazine
I enjoyed learning about the Dennett Group's visitor research for Southern Oregon. I want to follow up on the brief conversation we had about potential problems.
First consider the following simple model. Cities A and B each have 1,000 population. Everyone in A visits Southern Oregon once a year. 100 people in B visit Southern Oregon 10 times a year and no one else. If you sample the populations of A and B at home, you will decide that 10X as many SO visitors are from A as B. If you conduct intercepts in SO, you will get the same number of A and B. Longwoods is based on the former, and you on the latter. You won't get the same results.
You remarked that the only problems to worry about would be systematic sources of error. Here are five.
1) Assume that there is a decreasing likelihood on any particular trip of someone going to a Visitor Center. Your method will understate the number of trips to SO that a particular visitor takes.
2) However, since the cumulative likelihood of being at a visitor center at least once increases with more trips, you will paradoxically overweight the demographics of people who visit SO repeatedly.
3) You will underweight low incomes, because such people are less likely to base decisions on reading matter, which is what they get in visitor centers.
4) You will underweight "visiting friends and relatives" because such people have a built-in source of visitor information that is not available to FIT folks. They are consequently less likely to enter a visitor center.
5) You will underweight people who get their information from the Internet. Someone who has printed a local map from Mapquest doesn't need to get one from the visitor center.
You have a very accurate survey underway of exactly what it appears to be -- the profile of the average person who walks into a visitor center, or to a small degree an attraction. You also have useful information for the lodging, shopping, and dining folks, but it certainly won't be definitive.
I have devised a strategy for doing an online survey for the Oregon Coast. It will have errors, as does yours, but it will give useful information. It will say less about visitors to visitor centers but more about occupants of hotel rooms. I will be curious to see what differences appear between your results, my results, and Longwoods.
Rob Spooner, Publisher
Oregon Coast Magazine
