Calculating a mean is a statistical technique.

So you went to 2500 people and asked them if they were willing to pay more. 46% said yes, 32% said no, and the rest were undecided?

You could use something like the Excel macro called PooledInfRate (http://www.cdc.gov/westnile/resourcepages/mosqsurvsoft.html), recode the undecided as no and then calculate % yes along with a confidence interval. This will give a lower bound to the % yes if you can count on everyone who said yes to then follow through and vote yes.

Of course if the survey was more complex, then this might not be useful -- or it might just be a good start. Maybe the survey also asked how people felt about the increased fees. Then a "yes" vote could be weighted by feelings as a predictor of actual ballot response. Both in terms of how likely they are to vote yes, but also how likely they are to vote at all. The survey might show that 75% of people would vote yes, but if 68% of these people don't vote and 100% of the people against do vote, then the no vote wins.

Glen,

Thank you for that response. I should have but did not realize that Willingness to Pay was a specific set of concepts in economics. Of course, a simple search found a Wikipedia page that, while limited, helped. Unfortunately, things just got very complex.

So my willingness to pay is a function of:

My current income

My expectation of future income

How much I value the item/service in question

My valuation of other activities that I will have to sacrifice in order to pay more.

My valuation of external benefits.

Is this problem as simple as deciding if company X should increase the price of candy by 3 cents? Is this problem the willingness to pay increased utility costs to cover power company increased costs associated with implementing clean air regulations? Or is this something like willingness to pay more tithes to a local church which is done to improve my standing in my community? Or willingness to pay higher food prices when I am already missing meals.

The correct answer to the original question depends on who was surveyed and how they were surveyed. All we really know is that 2500 surveys are in hand. Were 250,000 surveys sent out, or just 2500? And there follows a long list of other questions.

Is there a statistician that you can talk to?javascript:

Use the binomial test (binom.test function in R)

My willingness to pay more is partly a function of how much more. It may be relatively easy to pay one cent more on a 20 dollar product, especially if I can see that future increases are unlikely. I am less likely to agree to pay 1020 dollars for the same product without considerable incentives. So willingness to pay is a continuous function determined by existing conditions, expectation of future conditions, and the amount of the increase. Willingness to pay is the maximum an individual will sacrifice to acquire or avoid something. It changes from person to person, so one is usually more interested in an average value and the rate at which that value changes with price.

That said, one could write the survey in such a why that the outcome is binomial. Ask a question, "would you be willing to pay a dollar and twenty three cents more for your cup of coffee?" However, this gets you a limited amount of information because you would expect the proportion of yes answers to change if you asked the same question but used $1.20 or $1.30 rather than $1.23. I would be suspicious of results from such a forced binomial model without some explanation. So I am an international producer of widgets. My widget costs $5.00. I want to raise the price by $0.01. However, for every penny I raise the cost, the consumer will spend $1.00 more in their currency. If I already know that they will not be willing to pay $2.00 more, then I can justify asking a simple binary response question.