Management of Non response in Market surveys

Refusals: Refusal rates vary from project to project and may range up to 25 percent. Since refusals are often the result of personality and mood, it can be argued that they occur randomly and will not bias results. Moreover, refusals are a matter of degree and of circumstances such as convenience at time of call. Repeated efforts to obtain compliance can reduce the refusal rate, but only to a degree. The older the respondent, the lower the social class; the larger the city, the more probable is a refusal.

In addition to general refusals, refusals may occur on specific questions, particularly those relating to income. When compared with people reporting income those refusing to report tend to include a larger percentage of upper income households (as inferred from rental data); highly educated persons; small families; older family heads; managerial and professional people; US born Protestants and Jews.

Refusal rates have shown a decided upwards trend since 1950, to 8 percent in small towns to 21 percent in metropolitan areas. These increases have been concurrent with the growth in use of random-digit dialing and have not been amendable to reduction even by the use of highly skilled interviewers and repeated callbacks. This problem appears to be related to the increased concern for safety and privacy that is particularly related to large cities.

Measurement of Non-response: A first step in coping with non-response is to develop accurate information on the problem. Surprisingly, standard measures of non-response (or response) have not been established in the field. To understand the non-response problems it is helpful to look at a ‘tree’ such as that shown below. This identifies the various things that can happen in a telephone survey and with small modifications, will also represent what can happen in a door-to-door survey.

A study of 40 different firms to find how response was measured turned up 29 different methods. Applying these methods to a single survey resulted in responses rates varying from 10 percent to 88 percent. The researchers presented data from a hypothetical survey and asked a group of research companies how they would calculate the response rate. The hypothetical survey was described as a telephone survey among a sample taken from telephone directories with the following results:

Category Frequency

Total numbers dialed 4,175
Discounted, nonworking numbers 426
No answers busy not at home 1,757
Interviewer reject (language barriers, deaf etc) 187
Household refusal 153
Respondent refusal 711
Ineligible respondent 366
Termination by respondent 74
Completed interviews 501

The three must frequently reported methods of calculating response (each given by only 7.5 percent of the firms) were the following:

Household refusals + Rejects + Ineligibles + Terminations + Refusals + Competed interviews
Total numbers dialed

153 + 187 + 366+ 74 + 711 + 501 = 48% response

Rejects + Ineligibles + Termination + Refusals + Completed interviews
Total numbers dialed

187+ 366+ 74+ 711+ 501 = 44% response

Completed interviews /Total numbers dialed = 501/4,175 =12% response

In the first reported calculation, the researchers are apparently defining responses as any contact made with the sample telephone number. The second calculation apparently considers response as a contact with the appropriate individuals in the house, and the third calculation defines response as a completed interview. The only way that methods one and two can be said to reflect true rate of response is to assume that those who were reached by phone, but with whom interviews wee to completed were the same as those with whom interviews were completed. As has been shown earlier, such an assumption is dangerous; one can well assume the no answers, busy signals and not-at-homes are similar to those with whom an interview was completed.