Surveys using random telephone number dialing, such as the Santé publique France Barometer, make it possible to reach virtually the entire population, since more than 99% of the population now has a landline or mobile phone (2021 ICT Survey, INSEE) (1). These surveys are fairly simple to implement since they involve dialing random telephone numbers. Over time, they have had to adapt to changes in telephone equipment and usage, particularly with the emergence and widespread adoption of mobile phones. Thus, in the early 1990s, phone numbers were selected at random from the phone book; later, landline numbers were dialed at random, with an incrementing method allowing for the calling of numbers not listed in the phone book. Mobile phones were gradually integrated and now account for more than half of the numbers called in this type of survey (60% to 70%). The question posed in this study is: “Can we do without calling landline numbers and switch to 100% mobile?” when conducting a general population survey in metropolitan France. To answer this, we conducted an analysis of the impact of this shift, both in terms of the characteristics of the surveyed population and in terms of estimates of key health indicators (perceived health status, physical inactivity, smoking, and suicidal behavior). The analysis is based on data from the 2017 Santé publique France Barometer: we compared the results obtained from the entire sample (≈25,000 people)—comprising a mix of landline and mobile numbers (the method used for the Barometer since 2014)—with the results that would be obtained using only data collected via mobile phone (≈15,000 people). This analysis of the impact of a change in survey methodology is particularly important for repeated surveys such as the Barometer, which study trends, and for which a break in the data series could be detrimental to the interpretations.

Conducting surveys via cell phone offers several advantages. First, in most cases, it is a personal device: interviewing the primary user of the randomly selected number therefore usually means interviewing the person who answers the call. In contrast, a landline number corresponds to a household, and a random selection must be made to interview someone within that household. Calling a mobile number therefore reduces the duration of the interview since this selection phase (about 2 minutes) is avoided, and also increases the accuracy of the estimates since the sampling plan is then a single-stage process (compared to a two-stage process for landlines). Furthermore, the number of calls required to interview a person is lower on a mobile phone, which allows the fieldwork to be completed more quickly: on average, 6 mobile phone numbers must be randomly selected and called to secure an interview, compared to 16 landline numbers. This significant difference is primarily due to a higher proportion of ineligible numbers (invalid numbers, businesses, etc.) on landlines. Finally, with mobile phones, the respondent chooses where to answer the call and where to continue the interview; this can allow for greater confidentiality and more candid responses.

The disadvantage of mobile phone numbers is that very few (1%) are listed in the phone book; therefore, we cannot send a notification letter to people to inform them of the interviewer’s call. This preliminary notification step is routinely performed for phone numbers listed in directories and helps encourage participation. However, this limitation is not specific to mobile phones, since fewer than 10% of landline numbers are listed in the phone book. Another drawback of calling a mobile phone is that the number is displayed, and the person may then decide not to answer if it is an unknown number. Thus, in 2021, 68% of mobile phone owners reported answering only when they recognized the caller’s number; this is the case for 48% of landline owners who systematically screen calls (2021 ICT Survey, INSEE) (1).

The article shows that conducting a survey using only mobile phone numbers makes it possible to reach all segments of the population and that the distribution of the sample of respondents would be as close to the population structure as in a survey comprising 60% mobile and 40% landline numbers, as is the case with the 2017 Santé publique France Barometer. We observe a few differences that are worth noting: the 100% mobile approach makes it easier to reach younger populations (under 35), those with higher education (high school diploma or higher), and urban residents, and conversely, it is more difficult to reach older adults (65–75 years old), those with lower levels of education, and people living in rural areas.

Regarding estimates of health behaviors, these are similar between the two types of surveys (100% mobile or 60% mobile/40% landline). For example, the proportion of people reporting limitations in daily activities is estimated at 21.6% in the 60% mobile/40% landline survey and 21.1% in the 100% mobile survey, while the proportion of sedentary individuals is estimated at 8.7% and 8.9%, respectively. Thus, switching to a 100% mobile survey would not radically alter the associated health messages. There are, however, some caveats: the estimate of daily smoking is higher in the 100% mobile survey (+0.6 percentage points for the general population, +1.0 points among 18-30-year-olds, and +1.1 points among 60-75-year-olds); the magnitude of the difference remains modest, but nevertheless corresponds to the year-over-year trend observed for this indicator. In other words, this difference may mask a favorable or unfavorable trend in health behavior.

The results of our study allow us to consider 100% mobile random-generation surveys for the agency’s future surveys: we have shown that this enables us to obtain a sample of respondents whose structure closely matches that of the population and to produce estimates of health behaviors with greater precision. This is a more effective solution that could be prioritized for new surveys, particularly among younger audiences. Regardless of the device used to make the call, what matters in this type of survey is the calling protocol, which must be persistent (involving multiple calls at various times) to ensure that hard-to-reach or less-willing-to-respond individuals are included among the respondents. These individuals—for example, those who agree to participate after initially hanging up before the interviewer could present the survey—may be less aware of health messages; for instance, they have heard less frequently about the “Month Without Tobacco” campaign, as shown by the results of the 2019 Public Health France Barometer (2). This is why it is important to combine high-quality sampling with a robust protocol: this ensures the quality of the survey and its estimates, and allows for an accurate representation of all situations.