The incidence of understanding incidence

Editor's note: Warren Pino is president of Q & A Research, Inc., Novato, Calif.

One of the most critical yet often confusing measures in marketing research is incidence. What is it? How exactly do you calculate it? And what is the impact of it on overall project costs?

Incidence, as it is used in marketing research, quite generally means the percentage of a population that has or does something in common. For example, the incidence of being left-handed is roughly 10 percent, for being a female it is 51 percent. As multiple "qualifiers" are a part of most surveys, such as women who are left-handed, the incidence figures are simply multiplied together to yield the "net incidence," which in this case would be 5.1 percent (.10 X .51).

Well, that was simple enough, except when you get a dialing tally from a supplier on a telephone study with 20 or even 30 separate call classification categories. With all this information, how do you know which ones are included in the incidence calculation and which aren't? Let's say you get the following tally after the first night's dialing:

	Category		#
	No answer		20
	Busy		10
	Answering machine		15
	Language barrier		10
	Initial refusal		10
	No woman in household		75
	Not left-handed		210
	Qualified terminates		5
	Complete interviews		10

Okay, so what's the incidence? Remember our definition of incidence here: the percentage of the population that has something in common. This implies a fraction, doesn't it? We want to put in the numerator all "qualifiers" and in the denominator we'll put all contacts.

What's a contact? It is a person that we can put into one of two buckets: qualified or unqualified. Everyone else is excluded from the calculation because we really don't know which bucket to put them in. For example, where do you put a no-answer or a busy? We have no idea whether they qualify or not, so we'll exclude them. The same goes for answering machine, language barrier and initial refusal. "No woman in household" goes into the unqualified bucket, as does "not left-handed." Qualified terminates are those respondents who initially qualified, but for one reason or another elected not to complete the interview. Along with the completed interviews, they'll go into the qualified bucket.

Okay, back to our original question. What's the incidence? I've calculated it as follows:

          5 + 10
-----------------------   =   5%
  210 + 75 + 5 + 10

Impact of incidence on costs
After the first day or two of dialing, unless there are quotas to be concerned with, using this formula should make it relatively easy to calculate your incidence. But exactly how does this impact your costs if your original estimate is off?

Let's assume that you anticipated that the incidence of qualification for a given study is 50 percent. Initial dialing efforts reveal that your assumption was off. In fact, way off. It has been calculated at 25 percent. If your supplier is doing its job, you should be given options to stay on budget (reducing the number of surveys, relaxing qualifying criteria, etc.) or a quote for additional costs based on the new incidence figure.

In this particular case, assuming no other specification changes, you should anticipate the field or data collection portion of your project costs to double. Why? Because it is twice as hard to find qualifiers with an incidence at 25 percent than it would have been at 50 percent. Further, it is important to remember that it is the relative difference in percentages between anticipated and actual and not the percentage difference alone that counts.

For example, let's now assume that we expect an incidence of 15 percent. Let's further assume that our actual rate of incidence ends up at 5 percent. While that is a spread of only 10 percentage points, it would actually be three times as hard to find qualifiers. The data collection costs should then be expected to triple the amount originally quoted.

Clients and suppliers alike need to know and appreciate the role that incidence plays in project costs. It's always a pleasant surprise when incidence assumptions are underestimated, but when they are overestimated it is important that everyone knows their options. If clients and suppliers agree on how incidence is defined and calculated, and understand the cost implications of differences between estimated and actual incidence figures, there will be fewer surprises and better client-supplier relationships.