Conjoint Measurement of Base Station Siting Preferences

ABSTRACT

Mobile communication has become a ubiquitous part of today's life. The ongoing growth of this technology, however, involves the construction of new mobile phone base stations in order to assure network coverage. The selection of a new base station site often results in conflicts between providers and public authorities, on the one hand, and residents on the other. The aim of the present study was to examine public preferences regarding base station sites. A random sample of 503 persons from the German speaking part of Switzerland was interviewed face-to-face in their homes. Conjoint analysis was used to evaluate participants’ preferences for various attributes of base stations (appearance, location, building, decision process). The results show that location plays the most important role in participants’ acceptance of base stations. The findings also indicate that most people would prefer a covered or camouflaged base station to a freely visible one. By means of a cluster analysis, several segments were distinguished, showing that base station siting preferences were not homogeneous. Implications for risk communication are discussed.

Key Words:

• electromagnetic fields
• conjoint analysis
• mobile communication
• siting

INTRODUCTION

Over the last decade, mobile communication has become a global phenomenon. Launched in Finland in 1991, mobile services increased to four billion mobile connections worldwide (ITU 2009). However, the ongoing growth of mobile communication and other technologies (Hilty et al. 2004) also gives rise to public concerns about potential health effects regarding electromagnetic fields (EMF). In particular, some citizens fear that mobile phone base stations may lead to severe health disorders, and strongly support corrective measures such as warning labels, appliance shielding, and further research (MacGregor et al. 1994).

From a technical point of view, in order to ensure network coverage, it is necessary to construct mobile phone base stations in the vicinity of places where people want to use their phones. The transmission power of base stations constructed outside of living areas is often not sufficient to ensure stable mobile reception; besides, the radiation from the mobile phone of a phoning person would be at a maximum. Thus, centrally located base stations reduce the mobile phone user's daily radiation exposure, because mobile phones and mobile phone base stations are connected entities: they radiate less when the distance between them decreases (Cousin and Siegrist 2010). Furthermore, because of the narrow vertical spread of the base station's beam, the levels of exposure inside or to the sides of buildings with base stations mounted on their rooftops are normally very low (WHO 2000).

In order to reduce public fears about EMF and to promote greater acceptance of a new facility, various courses of action have been suggested (Wiedemann and Schutz 2005). Among others, public participation in base station siting decisions is frequently mentioned by public authorities (e.g., WHO 2000). Little is known, however, about public preferences regarding mobile base station sites. It is unclear which characteristics of base stations are deemed most important by the public and whether there are considerable differences among segments of the population regarding site preferences. By means of a conjoint analysis, the present study is designed to address these issues in more detail.

Conjoint Analysis and Risk Perception

Conjoint analysis is a decompositional method developed to investigate preferences for hypothetical or real products and services. Respondents evaluate or rank a set of total profile descriptions and consider different attributes jointly, in contrast to compositional approaches, where various attributes are assessed separately (Green and Srinivasan 1978).

The biggest advantage of conjoint analysis is that it maintains a high degree of realism and resembles real choice situations (Hair et al. 1998). It is based on the assumption that it is easier for people to judge a product or service in its global utility than to assess each attribute in isolation. In addition, conjoint analysis is also useful in identifying different groups of respondents with similar preferences (Hair et al. 1998). These segments, usually derived using cluster analysis, can then be connected to various descriptor variables, such as sociodemographic or psychological variables. Despite its widespread use in other disciplines, especially in market research, only few studies used conjoint analysis to investigate risk assessment or environmental issues (Bond 2001; Gegax and Stanley 1997; Machado and Mourato 2002; McLean and Mundy 1998; Winslott Hiselius 2005). Winslott Hiselius (2005) investigated preferences for the transport of hazardous materials using a choice experiment, a variant of traditional conjoint analysis. In a similar vein, Machado and Mourato (2002) studied marine water quality improvements and health risk reductions. Choice experiments, however, do not allow for individual level analyses (Louviere 1991). To the best of our knowledge, no study has examined EMF risks using traditional conjoint analysis.

Identification of Relevant Attributes for Siting Decisions

For the design of a conjoint analysis, researchers should ensure that the attributes and attribute levels are both communicable and actionable (Hair et al. 1998). Beside these general characteristics, theoretical considerations on the identification of attributes should also be taken into account.

Several outcome variables (i.e., physical and tangible base station characteristics) may play a crucial role in base station siting preferences. First, the location of the base station might be a key issue, since research indicates that lay people often prefer great distance between themselves and the base station (Cousin and Siegrist 2010). Secondly, the appearance of the base station seems to be an important aspect as well. A study using a free association technique carried out by Siegrist et al. (2005a) suggests that lay people frequently mention aesthetic aspects when they think of mobile phone base stations. Changing the appearance of the base station, however, can be realized either by hiding the base station (behind sheeting, etc.), or by camouflaging the base station (e.g., as a tree or a crucifix). The latter option, in particular, might be perceived very differently because some might see a camouflaged base station as an aesthetic improvement, while others sense it as an attempt to deceive the public. Finally, a third outcome variable relevant for people's acceptance of base stations might be the type of building (WHO 2000). Residents often object to base stations in the vicinity of sensitive zones (e.g., kindergartens, schools or dwellings), although these sites would ensure minor exposure for the mobile phone users. Accordingly, placing a base station on a church might be perceived as inappropriate, or even as impious, by some people.

Besides these outcome variables, the process of the decision should also be taken into account. The question arises of whether people would rather reject or accept the idea that representatives from the public be allowed to participate in base station siting decisions. Several researchers have argued that one of the key features in environmental risk perception is fairness (Renn et al. 1995). Procedural fairness means that each person has the opportunity to express individual interests and can contribute to the collective will (Linnerrooth-Bayer 1995). However, some residents might also conclude that pubic representatives lack the knowledge to find appropriate sites, and thus reject this option.

The Present Study

Conjoint analysis is a fairly new approach in the field of environmental evaluation and permits a systematic understanding of preference structures. It has been suggested that it might be a useful tool for environmental risk analysis and communication (Alriksson and Oberg 2008).

After reviewing the relevant literature, we identified four key attributes to be crucial for base station siting decisions (i.e., the location, the appearance, the type of building, and the decision process itself). The purpose of the present work is to determine the relative importance of these attributes from the perspective of laypeople, and thus to explore which aspects of base station sites are considered most important by the public.

A second aim of the study is to detect specific segments of respondents that differ in their preferences. Some researchers have emphasized that individual differences in risk perception are often neglected (Siegrist et al. 2005c). Moreover, we wanted to examine if these groups of respondents would also differ in respect to other characteristics that were found to be important in the risk perception of mobile communication, such as trust (Poortinga and Pidgeon 2003; Siegrist et al. 2005b), knowledge (Cousin and Siegrist 2010), and health beliefs (Cousin and Siegrist 2008).

METHOD

Participants

The data were collected in a survey conducted in the urban area of Zurich, Switzerland. A sample of 503 participants aged from 18 to 80 was interviewed face-to-face in their homes between November 2008 and March 2009.

Participants were randomly selected from the electronic telephone directory. They were first contacted by mail and informed about the study. Approximately one week later, they were called by an interviewer and asked to participate in the study. The complete face-to-face interview lasted one hour on average. The response rate was about 35%. ¹ Interviewers were trained to conduct the interview in a standardized manner.

Forty-five percent (n = 225) of the respondents were female, and 55% (n = 275) were male. The mean age was 51.60 (SD = 15.10). Self-reported education level ranged from primary or lower secondary school (6%; n = 28), upper secondary vocational school or upper secondary university preparation school (59%; n = 294), to college or university (36%; n = 178). According to census data (BFS 2009), males were slightly overrepresented. Moreover, age and education level were slightly higher than the Swiss average. Ninety-four percent of the participants (n = 470) owned a mobile phone. On average, participants indicated that they use their mobile phone 26.04 (SD = 60.76) times per week for communication purposes. Three participants refrained from filling out the questionnaire that was presented at the end of the interview; thus, no demographic data were available for them.

Conjoint Analysis

In the conjoint analysis, ² hypothetical mobile phone base station sites were presented to the participants. The four attributes and their corresponding levels were: location (outside of the village/on the outskirts/in the center of village), appearance (visible/covered/camouflaged), building (factory/church/dwelling) and the decision process about the location of the base station site (residents were involved/government was involved/only provider decided).

Given that four attributes with three levels would yield 81 (3 × 3 × 3 × 3) possible combinations, it was necessary to reduce the number of the stimulus cards (i.e., the hypothetical mobile phone base station sites) by means of a fractional factorial design. The design and the number of the stimulus cards, which were presented in full-profiles, was determined through the ORTHPLAN procedure of the statistical software SPSS. The interviewers explained the meaning of the card attributes beforehand by means of an illustration that showed some additional explanations about the attributes. For example, the illustration showed three pictures of a mobile phone base station that was freely visible, covered with sheeting, or camouflaged as a crucifix. The specific attributes of the different sites were listed in table form and printed on cards (an example stimulus card is shown in Figure 1). The participants were asked to rank the cards according to their preference.

Figure 1 Sample stimulus card.

Questionnaire

The questionnaire was designed to measure a broad range of constructs that were related to mobile communication. For the present study, variables measuring the following five constructs were used: risk perception, benefit perception, knowledge, health beliefs, and trust.

Risk and benefit perception of mobile phone base stations was measured using the item: “How risky (beneficial) do you consider mobile phone base stations to be, for the Swiss society as a whole?” Participants responded on a 6-point scale ranging from 1 (“small”) to 6 (“large”).

The knowledge questions about mobile communication were taken from a scale developed by Cousin and Siegrist (2010). The original scale is divided into several domains. For the present study, the subscales “base stations” and “interaction patterns” were selected. The eleven items (e.g., “A base station gives off the same level of radiation throughout the whole day”) could be answered using the options “true,” wrong,” or “don't know.” A summative index of the correct items was calculated for further analysis. In our sample, the mean of the scale was 3.40 (SD = 2.48), indicating that participants had little knowledge about mobile communication (only one person was right on all of the eleven items).

Health beliefs were assessed by means of a scale introduced by Cousin and Siegrist (2008). The sixteen questions (e.g., “There are some people who can feel even low levels of radiation”) could be answered using the options “true,” wrong,” or “don't know,” and were summarized to a scale. High values on the health belief scale indicate a strong belief in adverse health effects of mobile communication. On average, participant agreed on 7.25 (SD = 3.40) items.

Finally, trust in mobile communication authorities was measured by asking participants to evaluate how much they trusted several authorities and their specific sphere of authority. They indicated their trust in (a) providers (technical aspects), (b) providers (health aspects), (c) federal authorities (legal framework), (d) federal authorities (health care), (e) research groups at universities (research), and (f) consumer protection boards (consumer safety). Participants answered these questions on a 6-point-scale ranging from 1 (“no trust”) and 6 (“complete trust”). For the final trust scale, a mean score of all items was calculated. The reliability of the scale was checked by calculating Cronbach's alpha (α = .66). Participants trusted research groups at universities most (M = 4.68, SD = 1.01), and providers (both items) least (M = 3.01, SD = 1.09).

RESULTS

Part-Worths and Relative Importance of Attributes

Based on the respondents’ ranking, part-worth utilities for each individual respondent and for the total sample were calculated (using SPSS CONJOINT). Part-worth utilities provide a quantitative measure of the preference for each attribute level (SPSS 2007). They were estimated using Ordinary Least Square (OLS) regression analysis and used to determine the relative importance values of each attribute. Importance values are calculated by taking the part-worth utility range for each factor separately and dividing it by the sum of the utility ranges for all factors. Thus, they provide a measure of how important the factor is to overall preference (SPSS 2007).

The attribute of greatest importance was the location of the mobile phone base station (35%). Decision process and building were of almost equal importance (24% and 23%, respectively), and the appearance of the base station had the lowest importance (17%). Thus, location was twice as important as appearance.

The part-worth utility scores for each attribute are shown in Figure 2. The results clearly showed positive utility for a location outside of the village; a location on the outskirts had only a slightly positive utility. In contrast, locations in the center of the village yielded a negative utility.

Figure 2 Part-worth utilities for each attribute level (all respondents).

Furthermore, the appearance of the mobile phone base station had a positive utility when it was covered or camouflaged. When the base station was visible, appearance resulted in a negative utility. Concerning the building, results demonstrate that participants preferred a mobile phone base station that was constructed on a factory, but they rejected a base station that was mounted on a dwelling. With regard to the decision process, a positive utility for a decision where residents were involved is indicated in Figure 2. Also, a decision process in which the government was involved had a positive utility, while a decision process where only the provider decided about the location of the base station clearly yielded a negative utility.

Segmentation

Cluster analysis was applied to classify participants into homogeneous subgroups. A K-Means clustering based on the individual part-worth utilities was used. A four-cluster solution showed high substantial interpretability, and clusters differed significantly on all variables used for clustering. Shown in Table 1 are the results of the cluster analysis, together with the size of each cluster.

Table 1 Utility of each attribute level, and importance of the attributes “location,” “appearance,” “building,” and “decision process” in groups obtained through cluster analysis.

	Cluster 1	Cluster 2	Cluster 3	Cluster 4
	(n = 54)	(n = 73)	(n = 167)	(n = 209)	F	P
Utility of location “outside of village”	.77	–.47	.54	2.42	266.27	.001
Utility of location “on the outskirts”	−.25	.17	.07	.33	10.41	.001
Utility of location “center of village”	−.52	.30	−.60	−2.75	416.74	.001
Utility of appearance “visible”	.20	−2.15	−.11	−.25	124.84	.001
Utility of appearance “covered”	.17	.89	.16	.18	19.69	.001
Utility of appearance “camouflaged”	−.37	1.26	−.05	.07	62.27	.001
Utility of building “factory”	1.57	.65	.89	.56	19.00	.001
Utility of building “church”	−1.91	−.16	.44	.14	121.37	.001
Utility of building “dwelling”	.34	−.49	−1.32	−.70	41.41	.001
Utility of decision process “residents”	.27	−.05	1.54	.46	71.65	.001
Utility of decision process “government”	−.05	.51	.58	−.03	25.59	.001
Utility of decision process “provider”	−.22	−.45	−2.12	−.43	156.92	.001
Importance of location	20.57	21.14	17.51	57.30
Importance of appearance	20.58	39.47	13.81	12.01
Importance of building	41.08	18.90	27.87	16.84
Importance of decision process	17.77	20.50	40.81	13.86

To describe the clusters, one-way analysis of variance (ANOVA) and Tukey-HSD post-hoc analyses were applied for the continuous variables (risk perception, benefit perception, knowledge, health beliefs, trust, and age). The results are presented in Table 2. Age did not reach significance and is therefore not presented in the table (p > .28). For the discrete variables (gender and education), a chi-square test was performed. Gender (χ = 8.06, p = .045), but not education (p > .20), was a significant descriptor for the four segments.

Table 2 Means (with standard deviations in parentheses) of group characteristics and statistical differences obtained via analysis of variance.

	Cluster 1	Cluster 2	Cluster 3	Cluster 4	F	P
Risk perception	3.59 (1.35)^b	2.79 (1.03)^a	3.58 (1.23)^b	3.63 (1.22)^b	5.46	.001
Benefit perception	4.11 (1.46)^a	4.95 (1.15)^b	4.10 (1.21)^a	4.17 (1.12)^a	9.331	.001
Knowledge	3.15 (2.65)^a	4.18 (2.90)^b	3.36 (2.48)^a,b	3.20 (2.23)^a	3.07	.028
Health beliefs	6.94 (3.72)^b	5.53 (2.55)^a	7.67 (3.49)^b	7.60 (3.34)^b	8.28	.001
Trust	3.94 (0.77)^a,b	4.05 (0.74)^b	3.72 (0.68)^a	3.76 (0.78)^a	4.15	.006
^a,bDifferent letters in a row indicate statistical difference, according to Tukey HSD post hoc analysis (p <.05).

The findings indicate that participants in cluster 1 (11%) place great importance on the type of building on which the base station is constructed. They would not accept a mobile phone base station on a church, and prefer that it be built on a factory. Cluster 1 is the smallest segment. It is the only segment that favors base stations that are freely visible and refuses camouflaged base stations. In contrast to cluster 2, it has less knowledge about mobile communication.

Participants in cluster 2 (15%) mainly attach importance to the appearance of the base station. They reject freely visible base stations and rather prefer one that is camouflaged or covered. Cluster 2 is the only group that prefers a base station that is located in the center of the village, which would be the best location from a public health perspective (Cousin and Siegrist 2010; WHO 2000). Location, however, was not of high importance. Compared to the other groups, participants of cluster 2 perceive lower risks and higher benefits from mobile communication, have more knowledge, fewer health beliefs, and have more trust in mobile communication authorities. Participants of cluster 2 are predominantly male (70%).

Participants belonging to cluster 3 (33%) are mainly interested in the decision process. A decision process in which only the provider is included is clearly rejected. Rather, they would prefer that residents are involved in the decision process. Some importance is also attached to the attribute “building.” Like all other segments, they prefer a base station that is mounted on a factory. They share with clusters 1 and 4 that they see higher risks and lesser benefits of mobile communication.

Cluster 4 is the largest cluster (42%). Participants of this segment stress the importance of the location, and hardly take other attributes into account. They strongly favor a base station that is located outside the village and refuse base stations located in the center. Like cluster 3, they have little trust in mobile communication authorities. Compared to cluster 2, they also have less knowledge.

DISCUSSION

The findings of this study give insight to public preferences for mobile phone base station sites. Results show that a covered base station that is mounted on a factory, and whose remote location is determined by residents, had the highest utility.

Location of the base station plays the most critical role in acceptance of base station sites. Participants clearly prefer base stations outside of the village, which is in line with Cousin and Siegrist (2010), who showed that people favor locations that are located as distant as possible.

Less importance is placed on the decision process. However, the findings indicate that respondents do not value decisions in which only the provider determines the location, probably because providers are perceived as less trustworthy than governmental institutions. Other researchers have suggested that fairness plays an important role for residents (Earle and Siegrist 2008; Renn et al. 1995), which is reflected in the fact that participants appreciate decisions where residents are involved.

The attribute “building” was equally important as the decision process. Factories are clearly favored as appropriate buildings for base stations. Presumably, participants have drawn the conclusion that workers in a factory would only be exposed during the day, while residents of a dwelling would also be exposed at nighttime. Also, participants may believe that factory workers are at risk anyway, and that the radiation exposure would only marginally contribute to the total risk of a factory worker.

Appearance was the least important attribute, only half as important as the location of the facility. According to WHO recommendations (WHO 2000), aesthetic aspects should be taken into account in siting base stations. In fact, our results show that visible base stations are rejected by most participants. Furthermore, for a small portion of participants (cluster 1), aesthetic modifications that aim at camouflaging base stations are clearly rejected, probably because participants fear that camouflaged base stations are deliberately intended to delude residents. Nonetheless, all respondents showed a positive utility for covered base stations. Consequently, if aesthetic modifications of mobile phone base stations are intended (WHO 2000), covering base stations (with sheeting, etc.) would reach the broadest consensus among the population.

The cluster analysis identified four segments with different preferences, indicating that the sample was not homogeneous in their siting preferences. Cluster 2 differed from the other clusters in regard to their preferences and also with respect to several descriptor variables. To minimize radiation for the mobile phone user, it is reasonable to place base stations at a central village location, and cluster 2 was the only cluster that favored this site. Compared to other respondents, they were characterized not only by little risk and higher benefit perception, higher trust, and fewer health beliefs, but also by more knowledge. This is in line with Cousin and Siegrist (2010), who demonstrated that knowledge gaps were related to respondents’ preferences regarding base station siting. For example, using a forced-choice task, they could show that respondents with little knowledge about interaction patterns between mobile phones and base stations preferred locations which would cause more exposure for the phoning population.

Following this line of argument, our research provides a guide for risk communication regarding EMF. At first, it highlights the importance of knowledge transfer. Our findings confirm that people generally have little knowledge about the functionality of mobile phone base stations and the interaction pattern of mobile phones and base stations, which emphasizes that provision of information should focus on imparting these facts. Secondly, because we found that people who select advantageous sites also have less negative health beliefs, the results highlight that it is advisable for public health authorities to communicate as clearly as possible about the functionality and health effects of mobile communication in order to reduce ambiguities. Third, our findings also show that trust is crucial for risk communication, since it was found to be a significant descriptor for the clusters. In order to foster trust, communication programs designed by public institutions may, for example, facilitate a two-way communication process in which citizens have the possibility of expressing their needs (as suggested by Renn and Levine 1991).

The present study was aimed at exploring the siting preferences of a randomly selected sample from the general population. Results are therefore limited to participants who were presumably moderately involved in the topic. For this reason, results of the conjoint analysis might be different for residents who are directly affected by a real base station conflict in their neighborhood, because highly involved participants may weight the attributes and attribute levels differently. Moreover, only the most important attributes of a base station could be addressed in the present study. From a theoretical point of view, it would have been possible to further subdivide some of the attribute levels. The appearance of a freely visible base stations, for example, could be broken down into height, style, color, and so on. Increasing the number of attribute levels, however, would also lead to a higher number of stimulus cards, and sorting too many cards might overstrain participants.

A conjoint analysis as employed in the present study is thought to be instrumental for further investigation of public preferences for mobile base station sites. Since the findings of our study underscore, among others, the importance of knowledge for choosing adequate locations, it might be interesting to incorporate conjoint analysis in an experimental setting. An experiment may shed more light into the question of how much and what kind of knowledge is needed in order to enable citizens to come to appropriate base station site decisions. Such an experiment could be combined with a thought listing technique (Brock 1967; Greenwald 1968), in which participants are asked to list everything about which he or she was thinking during the ranking of the conjoint cards. This technique may give insights about the reasons why one option is preferred over the other, which remains speculative in traditional conjoint analysis.

Finally, we believe that the implementation of more realistic measures such as conjoint analysis in risk perception must not be restrained exclusively to an EMF topic. Conjoint analysis was found to be a useful tool when applied to mobile phone base stations preferences, but it may also be applied to preferences regarding, for example, power plant sites, waste incineration plants and other environmental risks. Beyond this, the segmentation of participants into homogeneous subgroups permits the tailoring of information material to the specific characteristics of those groups. Thus, especially in combination with other techniques such as cluster analysis or experimental methods, we conclude that conjoint analysis allows for an improved understanding of environmental risk perception and communication.

ACKNOWLEDGMENT

This study was funded by the Swiss National Science Foundation as part of a National Research Programme on “Non-Ionising Radiation—Health and Environment” (NFP 57).

Notes

The response rate was calculated as: “Response rate” = (“number of completed interviews”)/(“number in sample-number not eligible”).

Previous to the conjoint analysis, participants conducted a free association task and a Single Category Implicit Association Task; results of these tasks are not reported here.