Additional WTP Settings
Because our WTP approach can require hundreds or thousand of sampling of competitive scenarios, for speed concerns we recommend using the Share of Preference simulation method, rather than Randomized First Choice. In our opinion, the practicality of speed would seem to counteract any benefits we might see involving RFC's correction for product similarity when computing WTP.
For WTP to be scaled properly in terms of monetary units, you first should have edited the Attribute Info from the home tab of the simulator. On the Attribute Info tab, you must click that your price attribute is Continuous, Is Price, and specify Level Values for your price attribute to reflect the monetary units shown to respondents. For example, if the first level of price was $3.75, then you should specify that its level value is 3.75.
From the My Scenario Settings tab, when you click the settings (gear) icon to the right of the Willingness to Pay range settings drop-down, you can access additional settings that govern your WTP run.
WTP Product: here you select the single product (typically your firm’s product) for which WTP will be estimated. By default, the product on the first row of the product specification grid is selected; but you can select a different product on a different row.
Reference Levels: here you specify which levels are the reference levels relative to which WTP will be estimated. For example, let’s imagine you are computing WTP for the Speed attribute, which had the following three levels:
Low speed |
N/A (reference level) |
Medium speed |
$50 (relative to Low speed) |
High speed |
$120 (relative to Low speed) |
In the example above, Low Speed is the reference level and the value in terms of WTP is estimated for the other two levels with respect to Low Speed.
By default, the software chooses the lowest preference level as the reference level. But, you can change the reference level if you wish. The method the algorithm uses to identify the lowest preference level is consistent with the Share of Preference (MNL) simulation approach: we treat levels within an attribute as if they were product alternatives in a choice set; we exponentiate the raw utilities for the levels within an attribute (the alternatives) and normalize them to sum to shares of preference summing to 100 within each attribute; we then average those preference scores across respondents to identify the lowest preference level within each attribute.
Sampling: this dialog controls what type of random sampling to use when estimating WTP. The default is to not do bootstrapping, as bootstrapping takes much longer to run. Bootstrap samples allow us to estimate confidence intervals. The more samples you draw, the more accurate the estimates of the WTP confidence intervals. Bootstrap sampling involves selecting a sample of respondents from the original sample using sampling with replacement. This means that if we have 500 respondent records in our analysis, we draw 500 respondents from that pool of records with replacement, such that each respondent can be represented multiple times in the bootstrap sample or not at all.
Bootstrap sampling allows us to compute confidence intervals around the WTP estimates. We compute the standard deviation of the WTP results across the bootstrap samples. Then, to form a 95% confidence interval, we take the median WTP across the bootstrap samples +/- 1.96 times the standard deviation. If you want to use different confidence intervals such as 90% or 99%, you can change the confidence interval setting back on the Home tab, using the Confidence Interval setting at the very right-hand side of the toolbar ribbon.
Number of Bootstrap Samples: (default 300). The more bootstrap samples you simulate, the more precise your estimate of the confidence interval will be. For quick-and-dirty work, 30 to 60 bootstrap samples will give you a rough estimate of the confidence interval. For more precision, we recommend 300 or more.
Competitive Sets per Bootstrap Sample: If you have specified that your WTP product and/or the competitors can take on ranges of levels in the WTP simulation scenario, this setting specifies how many random draws of scenarios will be used in the WTP estimation before before moving on to the next bootstrap sample. The default is 30. You will obtain more precise results (especially of the confidence interval) if you increase the number of sets per bootstrap sample beyond 30; but the estimation will take longer.
Random Seed: Algorithms that involve random draws (such as bootstrap sampling or random draws of competitors’ features) need an integer (called a Seed) to initialize the algorithm. If you use a different seed, you’ll get a slightly different result. The more bootstrap samples you use and the more random selections of competitors (Competitive Sets) within each bootstrap sample, the less the starting seed will matter to the results.
