Adoption of Protective Behaviours: Residents Response to City Smog …aging.fudan.edu.cn/Upload/2018-01-15-43f80be626-7478-40da-a1b1 … · including Beijing, Shenyang, Hefei, Tianjin

Adoption of ProtectiveBehaviours: Residents Responseto City Smog in Hefei, China

Peng Cheng*, Jiuchang Wei*,**, Dora Marinova**and Xiumei Guo***School of Management, University of Science and Technology of China, 96 Jinzhai Road, Hefei, AnhuiProvince 230026, China. E-mail: [email protected]**Curtin University Sustainability Policy Institute, Curtin University, Perth, Western Australia, Australia.E-mails: [email protected], [email protected], [email protected]

City smog outbreaks in many Chinese cities aroused public concern about health risksand environmental pollution in China. This study investigates protective behaviours ofthe citizens in response to city smog using protective action decision to understandtheir actions. A survey was conducted in Hefei, Anhui in 2015, after the city experi-enced severe smog events. The data from 429 respondents supported the developedmodel. Hazard-related attributes were the strongest predictors for willingness to adoptprotective behaviours. Public risk perception only positively affected self-protectiveactions. Information sources affected risk perception and increased the respondents’self-protective intentions. Older adults and respondents with lower educational levelswere more likely to adopt environmentally friendly actions. The findings’ theoreticaland practical implications are also discussed.

1 Introduction

In recent years, city smog has become one of themost serious environmental hazards confronting the

Chinese government and affecting the country’s citi-zens. Severe smog incidents occur on a regular basisin many large- and medium-sized Chinese cities,including Beijing, Shenyang, Hefei, Tianjin and Shanghai.The severity of the problem attracts much publicityand many critics argue that such crises show that theindustrial development of China has come at the costof environmental pollution (Albert & Xu, 2016; Arm-strong & Ke, 2013).Although people rarely panic in the face of emer-

gency situations (Mileti & Peek, 2000), city smogcauses anxiety among the Chinese public, with manycitizens expressing serious concerns about the airquality in their residential and working areas. Calls arebeing made for the attention of the relevant

authorities, including suggestions that the governmentshould take forceful measures to deal with the smog.The list of concerns expressed by the residents coverissues, such as the bad weather affecting their normallife and mental health; being affected by the city smog;wearing masks; and staying at home in times of smogbeing costly and inconvenient. The risk perception ofthe smog forces many citizens to take practical mea-sures, such as seeking refuge behind surgical masks.Not only Beijing households have had to make adjust-ments to the smog hazard; other cities with lowerlevels of air pollution have also begun to pay attentionto such incidents. Protective actions are seen as nec-essary by everybody affected.

Tourists and visitors have also become more cau-tious. Figure 1 shows the monthly particulate matterunder 2.5 micron (PM2.5) concentrations and themonthly number of visitors to the capital of Chinaduring the first three quarters of 2014. The curve

© 2017 John Wiley & Sons Ltd DOI: 10.1111/1468-5973.12148

Journal of Contingencies and Crisis Management Volume 25 Number 4 December 2017

associated with PM2.5 concentrations is opposite tothat of tourist arrivals. Many previous studies have dis-cussed the effects of city smog on tourism. Indeed,city smog could ‘cause potential visitors to changetheir travel plans’ (Zhang, Zhong, Xu, Wang & Dang,2015: 2397). Considering people’s reaction to smog, itis important to understand the factors that can influ-ence risk perception and individual protective beha-viours in response to this particular hazard.

Risk issues, the way they are perceived, the relianceof households on information sources and risk treat-ment have been studied for most hazards and somethreats by Burton (1993). The reaction of the publicto air pollution has also been discussed widely in theliterature (Bickerstaff & Walker, 1999, 2002). Differenttheories of risk perception and public reaction havebeen used (e.g., protection motivation theory andstage theories such as the transtheoretical model, seeMartin, Raish & Kent, 2008; general hazards copingtheory, see Laska, 1990; Han & Nigg, 2011), but nostudy has examined the risk perception of the increas-ingly severe Chinese city smog using the protectiveaction decision model (PADM). The PADM, which is a‘multistage model that is based on findings fromresearch on the responses of people to environmentalhazards and disasters’, ‘integrates the processing ofinformation derived from social and environmentalcues, with messages that social sources transmitthrough communication channels to those at risk’(Lindell & Perry, 2012). Paton and McCLure (2013:112) emphasize that PADM focusses ‘on how uncer-tainty influences both the search for information andthe nature and quality of protective action decision’.In this study, we investigate the reliance of citizens onsmog information sources associated with risk percep-tion and the determinants of them adopting protectivebehaviours in response to city smog.

The remainder of this article is organized as follows.Section 2 introduces a conceptual model of risk

perception and the protective behaviours of the publictowards smog. Section 3 describes the survey partici-pants, measures and analytical tools while Section 4reports the main results. A detailed discussion of theresults is provided in Section 5. Section 6 concludesthe study.

2 Literature review

The PADM has already been tested for many risks(Lindell & Perry, 2004). Previous studies have adoptedPADM to explain the process by which people per-ceive risk and take preventive measures to respond toit in relation to events, such as hurricanes, earth-quakes and floods (Lindell, Arlikatti & Prater, 2009;Huang, Li, Liu, Hu, Liu, Chen & Li, 2012a; Huang, Lin-dell, Prater, Wu & Siebeneck, 2012b; Trainor, Murray-Tuite, Edara, Fallah-Fini & Triantis, 2012; Terpstra &Lindell, 2013; Lindell, Huang, Wei & Samuelson,2016a; Lindell, Mumpower, Huang, Wu, Samuelson &Wei, 2016b). Originally proposed to explain how peo-ple respond to imminent disasters (Lindell & Perry,1992), PADM is based on the assumption that peoplereceive abundant risk information from outside chan-nels. Such information and the individuals’ pre-existingbeliefs derived from past experiences can affect theirperception of risk. The model also considers people’sperception about possible protective actions to dealwith the uncertainty in risk, especially the attributesbeing considered when searching for, selecting, andadopting appropriate protective actions (Lindell, Lu &Prater, 2005). A recent revised version of PADM (Lin-dell et al., 2012) considers the development of riskcommunication programmes, evacuation mode andthe long-term hazard adjustment adoptions. Consis-tent with the updated PADM, previous studies haveasked people questions about their situation to under-stand the environmental and social contexts, socialinformation about a risk, and perception of threatsand protective behaviours (Huang et al., 2012a,b; Lazo,Bostrom, Morss, Demuth & Lazrus, 2015; Terpstraet al., 2013; Wu, Lindell & Prater, 2015, Wei et al.,2015b, Wei et al., 2016). These questions include thefollowing (Lindell et al., 2005): Does a real risk exist?Is the provided information sufficient? What are thesources for such information? Is taking protectiveactions necessary? What is the best alternative fortaking protective actions?In this study, citizens face the risk of city smog. Their

reaction to this risk generally includes three stages.First, the citizens (who differ in their comprehensive-ness of understanding risk information because ofdemographic characteristics) receive smog informationfrom multiple channels (Lindell et al., 2012). Second,they perceive and assess the risk of city smog and natu-ral protective behaviours. Third, they decide to take

0

5

10

15

20

25

30

35

40

45

50

0

20

40

60

80

100

120

140

160

Con

cent

ratio

n(u

g/m

3 )

Tou

rist

Arr

ival

s(T

en th

ousa

nd p

eopl

e)

PM2.5 concentrations Tourist arrivals

Figure 1. Tourist arrivals and PM2.5 concentrations in Beijing2014. Source: National Bureau of Statistics of China and BeijingEnvironmental Protection Bureau.

Journal of Contingencies and Crisis ManagementVolume 25 Number 4 December 2017© 2017 John Wiley & Sons Ltd

Protective Behaviors Towards City Smog in China 245

protective actions to cope with the city smog. Thesethree stages are consistent with the revised PADM;hence, the model is applied in this study. Following thestructure of PADM, we use demographic factors asantecedent variables, along with the perception ofthreats, attributes which people consider when select-ing and adopting protective behaviours and behaviourintention to explain citizens’ reaction to city smog.Based on the existing literature, we propose four

hypotheses which relate to information sources, riskperception, perception of protective behaviours andprotective behaviours. They form the conceptualmodel shown in Figure 2 and are explained below.

2.1 Risk information sources

Awareness and knowledge of risk are usually notacquired through direct personal experiences (Chung,2011); a variety of information sources contribute tothe way in which people respond to risk. The sourceplays a fundamental role in transmitting risk informa-tion, which is based on the classic six-componentcommunication model of source–channel–message–re-ceiver–effect–feedback (Lindell et al., 2012). Sorensen(2000) reviewed studies on hazard information andwarning, and identified the major factors which affectwarning response. There is empirical support thatsource familiarity can increase the warning response.Sources range from various social groups to authori-ties, news media and peers (Lindell & Perry, 2004).Risk information is generally distributed evenly overthe entire population (Lindell et al., 2005). Reviewing23 studies, Lindell and Perry (2000a) examined theinfluence of the primary group and mass media asinformation sources and found that these two sourceswere significantly related to seismic adjustment. Lindellet al. (2005) extended the previous findings by consid-ering the reliance of people in hurricane evacuation

on three information sources, namely peers, localauthorities and mass media. Their results revealed thatdifferent source utilization led to different decision-making in hurricane evacuation. A number of studieshave explored the effects of demographic characteris-tics in relying on information sources (Wei et al.,2015a). For example, those who are older, more edu-cated and with a higher household income preferresorting to Internet information when they areexposed to health risks (Bowen, Meischke, Bush,Wooldridge, Robbins, Ludwig & Escamilla, 2003; Diaz,Griffith, Ng, Reinert, Friedmann & Moulton, 2002).

2.2 Risk perception

Risk perceptions are often used to examine how peo-ple recognize and identify risk; they include under-standing the probability of a risk happening at aparticular place during a particular time and the prob-ability of consequences from this risk (Meyer, Broad,Orlove & Petrovic, 2013; Mileti et al., 2000; Wu et al.,2015, Zhu et al., 2016). The perception of the risk isusually considered a central factor influencing thereaction of people towards it in PADM (Lindell et al.,2012). Researchers typically ask respondents a seriesof questions about different risks in order to under-stand people’s perception about them. Different ana-lytic methods are used to show how well a risk canbe perceived and what factors can influence the viewsof people (Trumbo & McComas, 2003).

Risk perception is often related to information pro-cessing. In the information flow in PADM, sources areimportant factors that can influence the predecisionprocesses and subsequently perceived risk (Lindellet al., 2012). The reliance of individuals on informa-tion sources and channels, such as authorities, mediaand peers, can trigger attention and interpretation,which are critical to their perception of the threat

Demographic

Characteris�cs

Environmentally friendly ac�ons

Self-protec�ve ac�ons

Risk Percep�on

Smog informa�on sources

Official Channel

Media Channel

Peers

Hazard-related a�ributes

Resource-related a�ributes

H1a:+

H1a:–

H1b

H2a:+ H2bH3

H4a

H4b

H4c

Figure 2. Conceptual model. Note: Each path is identified by the hypothesis with which it is associated. +: Positive effect. �: Negative effect.

Journal of Contingencies and Crisis ManagementVolume 25 Number 4 December 2017 © 2017 John Wiley & Sons Ltd

246 Peng Cheng, Jiuchang Wei, Dora Marinova and Xiumei Guo

(Wu et al., 2015). For example, Brenkert-Smith, Dick-inson, Champ and Flores (2013) found that informa-tion from expert and nonexpert sources were bothassociated with risk perception.

2.3 Perception of protective behaviours

Studies on the perception of protective behaviourshave shown two aspects that are important to people,namely perceived efficacy and perceived cost (Zaleskie-wicz, Piskorz & Borkowska, 2002). Perceived efficacycan be assessed by how well people feel that protec-tive behaviours can shield them from a risk. Perceivedcost is the influence of the perceived barriers, includingfinancial, on carrying out particular behaviours.According to Lindell and Perry (Lindell et al., 1992;Lindell et al., 2000a) and their line of research, the per-ception of protective behaviours can be defined as hav-ing hazard-related and resource-related attributes.Hazard-related attributes, such as perceived efficacyfor protecting people, property and useful for otherpurpose, are used to describe the relationship betweenthe hazard and the protective behaviours. Resource-related attributes, such as money, knowledge, time,skill, tool, efforts and cooperation with others,describe the relationship between the resource adjust-ment and the hazard (Lindell & Perry, 2004). As indi-cated previously, hazard-related and resource-relatedattributes are linked to behaviour response (Lindell &Prater, 2002; Lindell et al., 2009). Lindell and Whitney(2000b) also found that hazard-related attributes had asignificantly positive correlation with behaviourresponse, whereas resource-related attributes had anegative correlation with it.

2.4 Protective behaviours

The protective behaviours of individuals in responseto a risk differ substantially (Burton, 1993). For exam-ple, in the case of seismic hazards, people can adoptpre-impact adjustments, such as strapping water hea-ters, establishing supplies of bottled water and cannedfood and purchasing insurance (Lindell & Perry,2000a). Terpstra et al. (2013) reported flood protec-tive behaviours of individuals, which included prepar-ing an emergency kit (water, food, radio, etc.),searching for information about the flood, preparingan emergency plan, using sandbags and buying floodinsurance. Laska (1990) pointed out moving toanother city to avoid the threat of flood as anextreme action which people would not adoptbecause it could be expensive and unsettling.

Finding and examining behaviour patterns is worth-while. Pro-environmental behaviour (PEB) has beenwidely discussed in previous environmental studies(Clark, Kotchen & Moore, 2003; Karp, 1996; Scannell

& Gifford, 2010), and it can be viewed as a mixture oftwo types of protective actions: self-interest (e.g., topursue a strategy that minimizes one’s own healthrisk) and of concern for other people, the next gener-ation, other species or whole ecosystems (e.g., pre-venting city smog that may cause risks for the healthof others and reducing smog emission) (Bamberg &M€oser, 2007). In a city smog situation, the citizens’protective behaviour can also be classified into twocategories according to PEB: self-protective actionsand environmentally friendly actions. Citizens can takeactions to reduce exposure to city smog (e.g., wearmasks, stay home during smog, use air filters andmaintain clean sanitation), and they can also alter theirown lifestyle in order to reduce smog emissions (e.g.,reduce car use and save resources). The former canbe considered as self-protective actions and the lattercan be seen as environmentally friendly actions.Although of substantial personal importance, self-pro-tective actions have no direct impact on the environ-ment. On the contrary, the environmentally friendlyactions have a direct impact on the environment butare of smaller significance for self-protection (Tobler,Visschers & Siegrist, 2012).A number of factors have been shown to be linked

to protective behaviours (Lindell et al., 2000a, 2005).Previous studies examined the correlations betweenrisk perception and protective behaviours for manyrisk situations, such as seismic hazards (23 studiessummarized by Lindell et al. (2000a)), floods (Terpstraet al., 2013; Zaleskiewicz et al., 2002), hurricanes(Huang et al., 2012a,b; Lindell et al., 2016a,b), volcaniceruptions (Perry & Lindell, 2008), tornadoes (Trainoret al., 2015) and water contamination hazards (Lindellet al., 2016a,b). These studies conclude that risk per-ception has a significant effect on behaviour response.Other factors are not always as significant. For exam-ple, according to Kollmuss and Agyeman (2002), peo-ple who are more concerned about environmentalissues might not necessarily want to adopt environ-mentally friendly behaviours but may want to adoptself-protective actions. The evidence indicates that theadoption of protective behaviours is related to thehazard-related and resource-related attributes notedby Lindell and Perry (2004). Recent research has alsofound that hazard-related attributes have a strongereffect on the adoption intentions of protective beha-viours than resource-related attributes (Lindell et al.,2000b, 2002). Demographic factors are also consid-ered important in influencing protective behaviours.The literature review by Bish and Michie (2010) con-cluded that gender, ethnicity, educational level, work-ing status and marital status affected protectivebehaviours during a pandemic. These factors are alsowhat we explore further in the case of smog in a Chi-nese urban environment.

Journal of Contingencies and Crisis ManagementVolume 25 Number 4 December 2017© 2017 John Wiley & Sons Ltd

Protective Behaviors Towards City Smog in China 247

2.5 Hypotheses

Building on previous research, this study appliesPADM to detect the determinants of individual pro-tective behaviours in response to city smog in Chinaand tests four hypotheses. The conceptual model ispresented in Figure 2. The following hypotheses arepostulated:

• Hypothesis 1a (H1a): Hazard-related attributes willbe positively correlated with protective behaviours,whereas resource-related attributes will be nega-tively correlated with protective behaviours.

• Hypothesis 1b (H1b): Hazard-related attributes willbe more strongly correlated with protective beha-viours than resource-related attributes.

• Hypothesis 2a (H2a): Risk perceptions will be a sig-nificant positive predictor of protective behaviours,whereas they will be less strongly correlated withbehaviours than hazard-related and resource-relatedattributes.

• Hypothesis 2b (H2b): Risk perceptions will be morestrongly correlated with self-protective behavioursthan environmentally friendly behaviours.

• Hypothesis 3 (H3): A significant relationship willexist between the ratings of the reliance on infor-mation sources and the risk perception of citizens.

• Hypothesis 4a (H4a): Demographic characteristics(age, gender, education and income) will be signifi-cantly related to hazard-related and resource-related attributes.

• Hypothesis 4b (H4b): Demographic characteristics(age, gender, education and income) will be signifi-cantly related to smog information sources.

• Hypothesis 4c (H4c): Demographic characteristics(age, gender, education and income) will be signifi-cantly related to protective behaviours.

3 Methodology

3.1 Sample and data collection

The above hypotheses were tested using a survey onresidents in Hefei, Anhui Province of China. Hefei, thepolitical, economic and cultural capital of this land-locked central province (see Figure 3), has more than3 million residents in its inner boundaries (and 7 mil-lion in the greater region). Identified by the EconomistIntelligence Unit as one of the top 20 emerging pow-erhouses in China (Huang et al., 2012a,b), Hefei wasalso the fastest growing metropolitan economy in theworld in 2012 (Metro economies, 2012). Smog inHefei is a common occurrence because of its fossilfuel-based industrialization and geographical location,which is responsible for slow air movements carryingindustrial emissions. The December 2013 Eastern

China smog seriously affected Hefei (as well as Shang-hai, Tianjin and Nanjing), with levels of PM2.5 reaching500 micrograms per cubic metre, and disrupted theentire infrastructure and operation of the city, includ-ing roads, airports and schools (Zhang, 2014). Theresidents of this vast area along the Yangtze River suf-fered from heavy pollution and the National Meteoro-logical Center issued several consecutive yellow alertsfor smog and fog (Xinhua, 2014). This background setthe importance of understanding the perceptions ofpeople about smog and their potential responses tohazard situations through testing the aforementionedhypotheses.

The survey was based on a questionnaire and wasconducted with residents of Hefei in April 2014 afterthe outbreak of city smog during the winter andspring period of 2013–2014. Among the 74 Chinesecities most affected by air pollution in 2013, Hefeiranked No. 17, with an average of the maximum dailyPM2.5 level (micrograms per cubic metre) of 383(Tan, 2014). Hence, the Hefei population was a suit-able investigative sample for the risk perceptions asso-ciated with smog.

Formal surveys were conducted face to face inHefei. Part of the sample members was selected fromMaster of Business Administration (MBA) and Masterof Public Administration (MPA) students at the Univer-sity of Science and Technology of China and the HefeiUniversity of Technology. The remaining sample mem-bers were selected from five communities in Hefei.We used the help of nine colleagues who were trainedas interviewers for the survey process; they providedinstructions and helped respondents better understandthe questions in the survey. The respondents repre-sented a diverse population across demographic andsocioeconomic strata and were asked to fill out thequestionnaire on a voluntary basis. They returned thequestionnaires to the interviewer after completion.Anonymity was specifically assured, and small tokens(USB flash drives) were provided to the respondentsas an incentive to increase the response rate.

The survey commenced on 12 April 2014 and wascompleted within a month. We distributed 600printed copies of the questionnaire by hand to ensurea timely, efficient and scientific data collection. In total,429 of the 600 questionnaires (a response rate of71.5%) were identified as useful for data analysis. Thishigh response rate indicates the concern and willing-ness of people to engage with such an environmentalissue. The mean age of the respondents was26.27 years; the standard deviation was 6.7 (13.8%were younger than 20; 81.8% were between 20 and40; 4.4% were older than 40 years); 55.7% of the sam-ple were males. The mean education level of therespondents was college graduate, and their meanannual income was between ￥30,000 and ￥60,000

Journal of Contingencies and Crisis ManagementVolume 25 Number 4 December 2017 © 2017 John Wiley & Sons Ltd

248 Peng Cheng, Jiuchang Wei, Dora Marinova and Xiumei Guo

(between US$4,878 and US$9,756 as of 6 December2014 at xe.com). Given that this study was explora-tory, we did not aim for the sample to be statisticallyrepresentative of the Hefei population; nevertheless, itprovided good coverage of its socioeconomic groups.

3.2 Measures used

The questionnaire measured risk perception (Cron-bach’s=.81) by asking respondents to report their per-ception of city smog consequences on a scale of ‘notat all’ (=1) to ‘almost certain’ (=5) in relation to eightissues, namely personal physical health, personal men-tal health, development of children, personal or familylife, work efficiency, personal property, local economicdevelopment and government performance. The totalrisk perception was the average rating of all percep-tions about the consequences of city smog.

Hazard-related attributes (Cronbach’s=.81), resource-related attributes (Cronbach’s=.85) and adoption

intention of protective behaviours (Cronbach’s=.70)were measured in relation to six different smog pro-tective behaviours, namely wear a mask, stay at homein times of smog, buy air filters, maintain clean sanita-tion, reduce personal car use and save resources.According to Lindell et al. (2016a,b), each protectivebehaviour was rated on two hazard-related attributes(protect effectively from city smog and useful for pur-poses other than city smog), four resource-relatedattributes (costs a lot of money, requires specializedknowledge and skills, requires considerable effort andrequires significant cooperation from others) andadoption intention (it is something I would like to do).All measurements were taken on a five-point Likert-type scale from ‘not at all’ (=1) to ‘a large extent’(=5). The six protective behaviours were grouped intotwo, namely 1) self-protective behaviours, such aswear masks, stay at home during smog, use air filterand maintain clean sanitation and 2) environmentallyfriendly behaviours, such as reduce car use and save

Figure 3. The map of Hefei geographical location.

Journal of Contingencies and Crisis ManagementVolume 25 Number 4 December 2017© 2017 John Wiley & Sons Ltd

Protective Behaviors Towards City Smog in China 249

resources. The scores for the hazard-related attri-butes, resource-related attributes and adoption inten-tion of each risk treatment choice were created bycomputing the average for an attribute across themanner.According to Arlikatti, Lindell and Prater (2007),

the smog information sources (Cronbach’s=.85) weremeasured by asking the respondents to rate the chan-nels that they prefer to choose. Three informationsources were provided, namely (a) official source,including local government, provincial or national tele-vision station, local newspapers and local news media;(b) media source, including national news websites(such as Sina and Sohu) and social media (such asmicroblog); (c) peers, including friends, relatives,neighbours and coworkers, advertising columns andcommunity flyers. Each source was rated using a scaleof ‘not at all’ (=1) to ‘a large extent’ (=5). We alsomeasured experience by asking the respondentswhether they or a family member had ever becomesick from city smog.The demographic factors were represented by four

variables, that is age, gender, annual household incomeand education. Age was measured in years. Annualhousehold income was constructed with five categories,namely less than ￥30,000 (=1), ￥30,000–￥70,000(=2), ￥70,000–￥120,000 (=3), ￥120,000–￥200,000(=4) and over ￥200,000 (=5). Education was also afive-point variable, namely less than high school (=1),high school (=2), vocational school (=3), college gradu-ate (=4) and graduate/professional school (=5).

3.3 Data analyses

We conducted factor analysis to assess whether thehazard-related attributes were distinct from resource-related attributes. The results of the principal compo-nent analysis (PCA) with a varimax rotation yieldedtwo factors: the Kaiser–Meyer–Olkin measure of sam-pling adequacy was .815, and the chi-square value wassignificant (p<.001) explaining 83.79% of the varianceamong the attribute scales. The results supported theconvergent validity and discriminant validity of the haz-ard-related and resource-related attributes.We also conducted a PCA with a varimax rotation

to identify key factors to assess the validity of theinformation source items. The Kaiser–Meyer–Olkinmeasure of sampling adequacy was .819, and the chi-square value was significant (p=.00). All three factorsexplained approximately 77% of the total variance,which was sufficiently high to account for theobserved intercorrelations. From the PCA, the vari-ables were loaded onto the three factors, namely offi-cial source, media source and peers.We tested the correlation matrices from commu-

nity and student samples to determine whether the

interitem correlations were equal (Huang et al.,2012a,b; Lindell et al., 2016a,b). Box’s M test indicatedthat covariance matrices were equal (Box’s M=71.71,F=1.27, p>.05). Following Gnanadesikan et al. (1977),we also conducted a graphical homogeneity test [seeArlikatti et al. (2007) for an example]. The equiva-lence of the patterns of intercorrelations among therisk perception, information source hazard andresource-related attributes and adoption intention ofprotective behaviours was assessed taking theobtained value of each correlation for student sampleand plotting it against the corresponding value of thatcorrelation for community sample. For example, onedata point is defined by plotting the value of the cor-relation between risk perception and authorities forstudent sample on the x-axis and the correspondingvalue of the correlation between risk perception andauthorities for the community sample on the y-axis.So the total number of data points is equal to the dis-tinct correlation coefficients in the correlation matrixfor each sample: k(k�1)/2=10(9)/2=45. Figure 4 showsthat the cross-plot of interitem correlations for com-munity and student samples is approximately linearand has no obvious outliers, so we can ignored thedistinction between the two samples and created apooled correlation matrix.

4 Results

Table 1 presents the descriptive statistics and bivariatecorrelations. The shaded correlations in Table 1 arestatistically significant at p<.05. As the correlationbetween EnFri and SePro was higher than the bench-mark of .60, we conducted a multicollinearity test(Liu, Ke, Wei, Gu & Chen, 2010). Any possible multi-collinearity is not a problem if the tolerance value isgreater than .10 or the variance inflation factors (VIFs)are less than 10 (Neter, Kutner, Nachtsheim &Wasserman, 1996). In our study, the result indicatesthat the lowest tolerance value was .688, and thehighest VIF was 1.453. Accordingly, multicollinearity

–0.2

–0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

–0.2 0 0.2 0.4 0.6 0.8

Stud

ent s

ampl

e

Community sample

Figure 4. Cross-plot of interitem correlations for community andstudent samples.

Journal of Contingencies and Crisis ManagementVolume 25 Number 4 December 2017 © 2017 John Wiley & Sons Ltd

250 Peng Cheng, Jiuchang Wei, Dora Marinova and Xiumei Guo

did not appear to be a significant problem in our dataset.

Table 2 presents the regression model of the adop-tion intention of protective behaviours. The resultspartially support H1a. For both behaviours, the haz-ard-related attributes had positive and statistically sig-nificant effect on adoption intention. By contrast, the

regression coefficients of the resource-related attri-butes were not statistically significant for environmen-tally friendly behaviours or had the incorrect sign forself-protective behaviour. As the resource-relatedattributes yielded insignificant or positive regressioncoefficients, H1b was not supported. However, theregression coefficients of hazard-related attributeswere larger than those of resource-related attributes.Our results only partially supported H2a. The

regression coefficient of risk perception indicated aninsignificant effect on environmentally friendly beha-viours, which was contrary to H2a. Risk perceptionhad a positive and significant effect on self-protectivebehaviours. These results also supported H2b becauserisk perception yielded insignificant coefficients forenvironmentally friendly behaviours.The respondents reported their reliance on infor-

mation sources, and their risk perception was affectedby them. Specifically, the utilization of official source(r=.12), media source (r=.10) and peers (r=.11) wassignificantly related to risk perception. This result sup-ported H3.The regression results indicated a weak support for

H4a. Most of the correlations of hazard-related andresource-related attributes with demographic variables(13/16) were not significant. Hence, demographiccharacteristics only played a limited role in predictinghazard-related and resource-related attributes of pro-tective behaviours. The effects of demographic charac-teristics on the reliance on particular informationsource were tested with H4b with the results showingthat age was positively related to authority (r=.17) andnegatively related to media channel (r=�.10) whileeducation level had the opposite effect. Furthermore,females were more likely to rely on information frompeers (r=.15). The relationship between demographic

Table 1. Correlations Among the Variables

Variables Mean SD 1 2 3 4 5 6 7 8 9 10 11 12 13 14

1. Age 26.27 6.70 12. Gender 1.51 1.31 �.21 13. Education 4.08 .85 �.10 �.17 14. Income 2.56 1.17 .15 .00 .07 15. Authorities 3.31 1.08 .17 .04 �.18 �.08 16. Media 3.84 .99 �.10 .05 .12 �.07 .38 17. Peers 2.79 .96 �.08 .15 �.06 �.01 .34 .25 18. Risk Perception 3.76 .66 .10 .05 .03 .02 .12 .10 .11 19. HaEnFri 3.87 .55 .07 .08 �.05 .06 .20 .18 .02 .19 110. ReEnFri 2.56 .85 �.04 .03 �.11 �.03 .04 �.05 .08 .06 �.00 111. HaSePro 3.36 .58 .05 �.07 �.08 �.07 .31 .23 .21 .11 .38 .12 112. ReSePro 2.47 .65 .03 �.00 �.19 �.14 .16 .14 .01 .15 .02 .63 .32 113. EnFri 3.30 1.14 .16 �.01 �.14 �.05 .17 .18 .08 .03 .40 .05 .21 .20 114. SePro 3.24 .69 .08 �.05 �.12 �.05 .19 .10 .19 .36 .24 .15 .56 .32 .43 1

Notes: HaEnFri: Hazard-related attributes of environmentally friendly behaviours; ReEnFri: Resource-related attributes of environmentallyfriendly behaviours; HaSePro: Hazard-related attributes of self-protective behaviours; ReSePro: Resource-related attributes of self-protectivebehaviours; EnFri: adoption intention of environmentally friendly behaviours; SePro: adoption intention of self-protective behaviours.All r>.10 significant at p<.05. All the significant correlations are highlighted.

Table 2. Prediction of Protective Behaviours

Predictor

Step 1 Step 2

b SE b SE

Environmentally friendly behavioursAge .11* .01 .12** .01Gender �.04 .04Education �.10* .06 �.11* .06Income �.07 .04Authorities .04 .05Media .01 .06Peers .00 .06Risk perception .09 .08HaEnFri .37*** .06 .385*** .06ReEnFri .04 .03Adjust R2 .188 .182F 11.213*** 33.602***

Self-protective behavioursAge .03 .00Gender �.04 .02Education �.07 .03Income .00 .02Authorities .01 .03Media �.04 .03Peers .11** .03 .10** .03Risk perception .19*** .04 .19*** .04HaSePro .44*** .05 .45*** .05ReSePro .11** .04 .13** .04Adjust R2 .369 .369F 26.747*** 65.237***

Note: *, ** and *** indicate p<.05, p<.01 and p<.001, respectively.

Journal of Contingencies and Crisis ManagementVolume 25 Number 4 December 2017© 2017 John Wiley & Sons Ltd

Protective Behaviors Towards City Smog in China 251

characteristics and protective behaviour was testedwith H4c. From the correlations of demographic char-acteristics with environmentally friendly and self-pro-tective behaviours, three of eight were significant. Age(r=.16) was positively related to environmentallyfriendly actions, whereas education was negativelyrelated to environmentally friendly actions (r=�.14)and self-protective actions (r=�.12).Although the correlations reported in Table 1 pro-

vided some support for the model in Figure 2, certainunpredicted regression coefficients remained. Forinstance, relying on information from peers (b=.11)had a direct significantly positive effect on self-protec-tive behaviours that were not hypothesized by Fig-ure 2. The results could be interpreted as indicationsof partial mediation, in which information sourcesexerted effects on self-protective behavioural intentionthat was not captured by risk perception.We conducted another regression analysis to test

the mediation effect of information source. As indi-cated in Table 3, regressing risk perception onto infor-mation source and demographic variables showed thatonly age had a direct effect on risk perception.Although significant correlations existed between riskperception and information source, the regressioncoefficients had no significant effect.

5 Discussion

Based on previous literature, we distinguishedbetween environmentally friendly behaviours and self-protective behaviours for city smog (Kollmuss et al.,2002). For the two types of protective behaviours, weexamined which factors could predict the intention ofcitizens to adopt protective behaviours to address citysmog.Our findings supported H1a, that is hazard-related

attributes should be positively correlated with protec-tive behaviours, whereas resource-related attributesshould be negatively correlated with them (H1a).However, contrary to hazard-related attributes,

resource-related attributes yielded insignificant corre-lations with environmentally friendly actions and posi-tive correlations with self-protective actions. Theseresults were not consistent with our hypothesis, andno clear reason was found. However, the results areconsistent with the findings of Lindell et al. (2002) andTerpstra et al. (2013) in the case of seismic and floodadjustment adoption intentions, respectively. Theseresults still support the view that hazard-related attri-butes would be more strongly correlated with adop-tion intention than resource-related attributes (H1b)according to the regression coefficients despite thepositive correlations of resource-related attributes.This finding could indicate that citizens were con-cerned with the state of the city smog and placedmore weight on protecting themselves than on per-sonal costs (e.g., loss of time, financial costs or incon-venience). This finding also reflects the cognitivedissonance of citizen (Festinger, 1962), that is peoplemight engage in high-cost behaviours that were actu-ally not effective. Moser (2007) found that climatebenefit could not be perceived because the connec-tion between the current action and its effects on theclimate was difficult to estimate. Whitmarsh (2009)also suggested that individuals were more likely tooverestimate the contribution of their behaviourswhile underestimating the negative impact of theiractions. Thus, information on hazard-related andresource-related attributes could help people reducethe cognitive dissonance and give them a feeling ofcontrol.

Risk perception was shown to be a significant andimportant factor only for self-protective behaviours.Our findings were in line with past research showingthat citizens with high level of risk perception foundself-protective actions more acceptable than actionsthat would change their lifestyle (Kollmuss et al.,2002). Poortinga, Steg, Vlek and Wiersma (2003) evenfound that people with high-risk perception about theenvironment were more willing to accept smallenergy-saving actions than actions that might result inlarge energy savings. The results also showed that haz-ard-related attributes were more predictive of protec-tive behaviours than risk perception. According toBish et al. (2010), perceived risk and belief in theeffectiveness of protective actions were the main fac-tors that influenced the adoption intentions of risktreatment. The results were also consistent with thetheory of reasoned action (Fishbein & Ajzen, 2011)and demonstrated that the attitude of people towardsadopting a risk treatment was more important in pre-dicting their behaviour than their attitudes towardsthe risk itself.

Significant correlations existed between risk percep-tion and the reliance of people on information source,although the regression coefficients were not

Table 3. Prediction of Risk Perception

Predictor

Step 1 Step 2

b SE b SE

Age .12* .01 .10** .01Gender .07 .02Education .07 .04Income .00 .03Authorities .06 .03Media .05 .04Peers .08 .04Adjust R2 .022 .008F 2.440*** 4.482***

Note: *, ** and *** indicate p<.05, p<.01 and p<.001, respectively.

Journal of Contingencies and Crisis ManagementVolume 25 Number 4 December 2017 © 2017 John Wiley & Sons Ltd

252 Peng Cheng, Jiuchang Wei, Dora Marinova and Xiumei Guo

significant. When city smog frequently occurs, citizensmight perceive a great risk because of the large infor-mation from outside and the increased uncertainty inthe risk. This finding was consistent with the conclu-sion from past research in other risk situations (Lin-dell & Hwang, 2008; Wei, Zhao, Wang & Zhao, 2016).Most people relied on the media channel to obtainsmog information although the influence of this infor-mation source on risk perception and risk treatmentadoption was smaller than that of the official channeland some peers. A possible explanation is that whenindividuals receive information from peers and theauthorities, they usually associate its purpose withmaking a warning. This finding is also supported byprevious studies (Lindell et al., 2005), which havereported that hurricane evacuation decisions can bepredicted better by peers and local authorities thanlocal media. Our study extended this finding byreporting that risk perception was also more stronglylinked to the official information sources and peersthan to the media channel.

Although the connections between demographicfactors and protective behaviours have been discussedin many studies, our hypothesis received only partialsupport. In this study, the demographic characteristicsof the respondents had an influence on their decisionsto select environmentally friendly behaviours but noimpact on their self-protective actions. Older respon-dents were more likely to avoid cars and saveresources, which is consistent with the previous stud-ies conducted by Tobler et al. (2012). This finding canbe attributed to the fact that elderly people in Chinahave a different lifestyle than the younger ones, includ-ing using mobile phones minimally and adopting amore frugal behaviour (Chen, Wang & Steemers,2013). Education was negatively related to the adop-tion intention of environmentally friendly behaviours,and this result was also consistent with the study byTobler et al. (2012). A possible explanation is thatqualified jobs often require more travelling to meet-ings using a car. Unpredicted direct links were identi-fied between age and risk perception, and this findingis consistent with studies which indicate that olderpeople are more willing to perceive higher risks(Rubin, Amlot, Page & Wessely, 2009). Older andpoorly educated people tended to rely on officialchannels, whereas younger and highly educated peoplepreferred to obtain smog information from the mediachannels. Women were more likely to talk with peersto obtain useful information. These results also sup-port previous studies (e.g., Cotten & Gupta, 2004).

6 Conclusion

This study adopted PADM to explain how people per-ceive risk and take preventive measures to respond to

city smog, as well as identified the determinants of theprotective behaviours of people in Hefei. Residentswere asked about their perception of protective beha-viours, perceived risk, reliance on information sourcesand demographic characteristics. The results revealedthat although risk perception influenced self-protectivebehaviours, it was not the most influential factorencouraging people to address city smog. By contrast,the hazard-related attributes of protective behaviourswere the strongest predictor of the willingness of peo-ple to engage with behavioural response. Demographicvariables and information source could also affect therisk perception and intention of the behaviourresponse of people.These results have implications for risk communica-

tion. They provide key insights into explaining protec-tive behaviours in response to city smog. In order toimprove community preparedness to deal with smog,the government, policymakers and managers shouldunderstand that people’s perceptions about behaviouralbenefits are more important than other factors. Speci-fying and emphasizing the usefulness and effectivenessof both types of behaviours to deal with city smogshould be considered by policymakers and managers.In line with the suggestions by Moser and Dilling(2004) and Ngo, West and Calkins (2009), emphasizingthe benefit of environmentally friendly behaviours (e.g.,reducing car use and using public transport) might beparticularly promising because it represents a positiveform of communication, which can effectively reducecity smog emissions and facilitate government manage-ment (Lu & Xue, 2016; Steg, Bolderdijk, Keizer &Perlaviciute, 2014; Steg & Vlek, 2009; Whitmarsh,2009). Accordingly, specifying the benefits of self-pro-tective behaviours can also help the public be moremindful of smog treatment, which makes them takemeasures to protect themselves efficiently (Burns &Slovic, 2012). Although resource-related attributes hadan insignificant effect on environmentally friendly beha-viours, it is important to indicate that they are associ-ated with hazard-related attributes. If these protectiveactions involve considerable time and effort or highcost, effective communication can help people realisti-cally estimate the consequences of their actions, par-ticularly to avoid misunderstanding and enable them toset priorities in selecting protective actions. For exam-ple, citizens can be reminded that maintaining cleansanitation may be more effective than spending moremoney on buying an air filter. Considering that infor-mation source is another important predictor, selectingappropriate channels to release useful smog informa-tion is important (Brenkert-Smith et al., 2013; Ver-roen, Gutteling & Vries, 2013; Wachinger, Renn, Begg& Kuhlicke, 2013). The findings of this study suggestthat authority and peer sources may be effective inchanging risk perceptions and motivating people to

Journal of Contingencies and Crisis ManagementVolume 25 Number 4 December 2017© 2017 John Wiley & Sons Ltd

Protective Behaviors Towards City Smog in China 253

adopt protective behaviours (Terpstra, Lindell & Gut-teling, 2009). City executives can also use bulletinboards in the community to broadcast appropriatesocial responses to city smog, considering that thishelps draw the attention of citizens and guides thepublic in responding to the situation.This exploratory study was the first to use PADM

to generate useful insights about the information reli-ance, perception and protective behaviours of people.Further research, including using a comparativeapproach, can provide additional evidence necessaryto obtain a good understanding about how peoplereact to the risk of city smog as well as what they areprepared to do in response.

Acknowledgement

The first and the second authors acknowledge thefinancial support from the Major/Innovative Programof Development Foundation of Hefei Center for Physi-cal Science and Technology (2014FXCX004), theNational Social Science Fund of China (14BGL141),and the Natural Science Foundation of China(71373250, 71522013 and 71490735). The third andforth authors also acknowledge the support from theAustralian Research Council. Finally, the constructivecomments from the Journal Editor and reviewershelped improve the quality of the manuscript forwhich we are very grateful.

ReferencesAlbert, E. and Xu, B. (2016), China’s environmental crisis, http://www.cfr.org/china/chinas-environmental-crisis/p12608 (accessed 26 April 2016).

Arlikatti, S., Lindell, M.K. and Prater, C.S. (2007), ‘Perceived Stakeholder RoleRelationships and Adoption of Seismic Hazard Adjustments’, InternationalJournal of Mass Emergencies and Disasters, Volume 25, Number 3, p. 218.

Armstrong, P. and Ke, F. (2013), Beijing announces emergency measures amid fogof pollution, http://edition.cnn.com/2013/10/23/world/asia/china-beijing-smog-emergency-measures/ (accessed 26 April 2016).

Bamberg, S. and M€oser, G. (2007), ‘Twenty years after Hines, Hungerford, andTomera: A new meta-analysis of psycho-social determinants of pro-environmental behaviour’, Journal of environmental psychology, Volume 27,Number 1, pp. 14–25.

Bickerstaff, K. and Walker, G. (1999), ‘Clearing the Smog? Public Responsesto Air-Quality Information’, Local Environment, Volume 4, Number 3, pp.279–294.

Bickerstaff, K. and Walker, G. (2002), ‘Risk, Responsibility, and Blame: AnAnalysis Of Vocabularies Of Motive in Air-Pollution (ing) Discourses’,Environment and Planning A, Volume 34, Number 12, pp. 2175–2192.

Bish, A. and Michie, S. (2010), ‘Demographic and Attitudinal Determinants ofProtective Behaviours During a Pandemic: A Review’, British Journal ofHealth Psychology, Volume 15, Number 4, pp. 797–824.

Bowen, D., Meischke, H., Bush, N., Wooldridge, J., Robbins, R., Ludwig, A.and Escamilla, G. (2003), ‘Predictors of Women’s Internet Access andInternet Health Seeking’, Health Care for Women International, Volume 24,Number 10, pp. 940–951.

Brenkert-Smith, H., Dickinson, K.L., Champ, P.A. and Flores, N. (2013), ‘SocialAmplification of Wildfire Risk: The Role of Social Interactionsand Information Sources’, Risk Analysis, Volume 33, Number 5, pp. 800–817.

Burns, W.J. and Slovic, P. (2012), ‘Risk Perception and Behaviors: Anticipatingand Responding to Crises’, Risk Analysis, Volume 32, Number 4, pp. 579–582.

Burton, I. (1993), The Environment as Hazard. Guilford Press, New York.

Chen, J., Wang, X. and Steemers, K. (2013), ‘A Statistical Analysis of aResidential Energy Consumption Survey Study in Hangzhou, China’, Energyand Buildings, Volume 66, pp. 193–202.

Chung, I.J. (2011), ‘Social Amplification of Risk in the Internet Environment’,Risk Analysis, Volume 31, Number 12, pp. 1883–1896.

Clark, C.F., Kotchen, M.J. and Moore, M.R. (2003), ‘Internal and ExternalInfluences on Pro-Environmental Behavior: Participation in a GreenElectricity Program’, Journal of Environmental Psychology, Volume 23, Number3, pp. 237–246.

Cotten, S.R. and Gupta, S.S. (2004), ‘Characteristics of Online and OfflineHealth Information Seekers and Factors That Discriminate Between Them’,Social Science & Medicine, Volume 59, Number 9, pp. 1795–1806.

Diaz, J.A., Griffith, R.A., Ng, J.J., Reinert, S.E., Friedmann, P.D. and Moulton,A.W. (2002), ‘Patients’ Use of the Internet for Medical Information’, Journalof General Internal Medicine, Volume 17, Number 3, pp. 180–185.

Festinger, L. (1962), A Theory of Cognitive Dissonance. Stanford UniversityPress, Stanford.

Fishbein, M. and Ajzen, I. (2011), Predicting and Changing Behavior: TheReasoned Action Approach. Psychology Press, New York.

Gnanadesikan, R., Kettenring, J.R. and Landwehr, J.M. (1977), ‘Interpreting andassessing the results of cluster analyses’, Bulletin of the International StatisticalInstitute, Volume 47, Number 2, pp. 451–463.

Han, Z. and Nigg, J. (2011), ‘The Influences of Business and Decision Makers’Characteristics on Disaster Preparedness – A Study on the 1989 LomaPrieta Earthquake’, International Journal of Disaster Risk Science, Volume 2,Number 4, pp. 22–31.

Huang, Q., Li, M., Liu, Y., Hu, W., Liu, M., Chen, Z. and Li, F. (2012a), ‘UsingConstruction Expansion Regulation Zones to Manage Urban Growth inHefei City, China’, Journal of Urban Planning and Development, Volume 139,Number 1, pp. 62–69.

Huang, S.-K., Lindell, M.K., Prater, C.S., Wu, H.-C. and Siebeneck, L.K.(2012b), ‘Household Evacuation Decision Making in Response to HurricaneIke’, Natural Hazards Review, Volume 13, Number 4, pp. 283–296.

Karp, D.G. (1996), ‘Values and Their Effect on Pro-Environmental Behavior’,Environment and Behavior, Volume 28, Number 1, pp. 111–133.

Kollmuss, A. and Agyeman, J. (2002), ‘Mind the Gap: Why do People ActEnvironmentally and What Are the Barriers to Pro-EnvironmentalBehavior?’, Environmental Education Research, Volume 8, Number 3, pp.239–260.

Laska, S.B. (1990), ‘Homeowner Adaptation to Flooding An Application of theGeneral Hazards Coping Theory’, Environment and Behavior, Volume 22,Number 3, pp. 320–357.

Lazo, J.K., Bostrom, A., Morss, R.E., Demuth, J.L. and Lazrus, H. (2015),‘Factors Affecting Hurricane Evacuation Intentions’, Risk analysis, Volume35, Number 10, pp. 1837–1857.

Lindell, M.K. and Perry, R.W. (1992), Behavioral Foundations of CommunityEmergency Planning. Hemisphere Publishing Corp, Washington DC.

Lindell, M.K. and Perry, R.W. (2000a), ‘Household Adjustment to EarthquakeHazard a Review of Research’, Environment and Behavior, Volume 32,Number 4, pp. 461–501.

Lindell, M.K. and Whitney, D.J. (2000b), ‘Correlates of Household SeismicHazard Adjustment Adoption’, Risk Analysis, Volume 20, Number 1, pp. 13–26.

Lindell, M.K. and Prater, C.S. (2002), ‘Risk Area Residents’ Perceptions andAdoption of Seismic Hazard Adjustments1’, Journal of Applied SocialPsychology, Volume 32, Number 11, pp. 2377–2392.

Lindell, M.K., Lu, J.-C. and Prater, C.S. (2005), ‘Household Decision Makingand Evacuation in Response to Hurricane Lili’, Natural Hazards Review,Volume 6, Number 4, pp. 171–179.

Lindell, M.K. and Hwang, S.N. (2008), ‘Households’ Perceived Personal Riskand Responses in a Multihazard Environment’, Risk Analysis, Volume 28,Number 2, pp. 539–556.

Lindell, M.K., Arlikatti, S. and Prater, C.S. (2009), ‘Why People Do What TheyDo to Protect Against Earthquake Risk: Perceptions of Hazard AdjustmentAttributes’, Risk Analysis, Volume 29, Number 8, pp. 1072–1088.

Lindell, M.K. and Perry, R.W. (2012), ‘The Protective Action Decision Model:Theoretical Modifications and Additional Evidence’, Risk Analysis, Volume32, Number 4, pp. 616–632.

Lindell, M.K., Huang, S.-K., Wei, H.-L. and Samuelson, C.D. (2016a),‘Perceptions and Expected Immediate Reactions to Tornado WarningPolygons’, Natural Hazards, Volume 80, Number 1, pp. 683–707.

Lindell, M.K., Mumpower, J.L., Huang, S.-K., Wu, H.-C., Samuelson, C.D. andWei, H.-L. (2016b), ‘Perceptions of Protective Actions for a WaterContamination Emergency’, Journal of Risk Research, http://dx.doi.org/10.1080/13669877.2015.1121906.

Journal of Contingencies and Crisis ManagementVolume 25 Number 4 December 2017 © 2017 John Wiley & Sons Ltd

254 Peng Cheng, Jiuchang Wei, Dora Marinova and Xiumei Guo

Liu, H., Ke, W., Wei, K.K., Gu, J. and Chen, H. (2010), ‘The Role ofInstitutional Pressures and Organizational Culture in the Firm’s Intention toAdopt Internet-Enabled Supply Chain Management Systems’, Journal ofOperations Management, Volume 28, Number 5, pp. 372–384.

Lu, X. and Xue, L. (2016), ‘Managing the Unexpected: Sense-Making in theChinese Emergency Management System’, Public Administration, Volume 94,Number 2, pp. 414–429.

Martin, W.E., Raish, C. and Kent, B.M. (2008), Wildfire Risk: Human Perceptionsand Management Implications. Earthscan Publications Ltd, London.

Metro economies (2012), The Economist, http://www.economist.com/node/21567579 (accessed 6 December 2014).

Meyer, R., Broad, K., Orlove, B. and Petrovic, N. (2013), ‘Dynamic Simulationas an Approach to Understanding Hurricane Risk Response: Insightsfrom the Stormview Lab’, Risk analysis, Volume 33, Number 8, pp. 1532–1552.

Mileti, D.S. and Peek, L. (2000), ‘The Social Psychology of Public Response toWarnings of a Nuclear Power Plant Accident’, Journal of HazardousMaterials, Volume 75, Number 2, pp. 181–194.

Moser, S.C. and Dilling, L. (2004), ‘Making Climate Hot’, Environment:Science and Policy for Sustainable Development, Volume 46, Number 10, pp.32–46.

Moser, S.C. (2007), ‘In the long shadows of inaction: The quiet building of aclimate protection movement in the United States’, Global EnvironmentalPolitics, Volume 7, Number 2, pp. 124–144.

Neter, J., Kutner, M.H., Nachtsheim, C.J. and Wasserman, W. (1996), AppliedLinear Statistical Models. Irwin, Chicago.

Ngo, A.T., West, G.E. and Calkins, P.H. (2009), ‘Determinants ofEnvironmentally Responsible Behaviours for Greenhouse Gas Reduction’,International Journal of Consumer Studies, Volume 33, Number 2, pp. 151–161.

Paton, D. and McClure, J. (2013), Preparing for Disaster: Building Household andCommunity Capacity. Charles C Thomas, Springfield.

Perry, R.W. and Lindell, M.K. (2008), ‘Volcanic Risk Perception andAdjustment in a Multi-Hazard Environment’, Journal of Volcanology andGeothermal Research, Volume 172, Number 3, pp. 170–178.

Poortinga, W., Steg, L., Vlek, C. and Wiersma, G. (2003), ‘HouseholdPreferences for Energy-Saving Measures: A Conjoint Analysis’, Journal ofEconomic Psychology, Volume 24, Number 1, pp. 49–64.

Rubin, G.J., Amlot, R., Page, L. and Wessely, S. (2009), ‘Public Perceptions,Anxiety, and Behaviour Change in Relation to the Swine Flu Outbreak:Cross Sectional Telephone Survey’, British Medical Journal (BMJ), Volume339, p. b2651.

Scannell, L. and Gifford, R. (2010), ‘The Relations Between Natural and CivicPlace Attachment and Pro-Environmental Behavior’, Journal of EnvironmentalPsychology, Volume 30, Number 3, pp. 289–297.

Sorensen, J.H. (2000), ‘Hazard Warning Systems: Review of 20 Years ofProgress’, Natural Hazards Review, Volume 1, Number 2, pp. 119–125.

Steg, L. and Vlek, C. (2009), ‘Encouraging Pro-Environmental Behaviour: AnIntegrative Review and Research Agenda’, Journal of EnvironmentalPsychology, Volume 29, Number 3, pp. 309–317.

Steg, L., Bolderdijk, J.W., Keizer, K. and Perlaviciute, G. (2014), ‘An IntegratedFramework for Encouraging Pro-Environmental Behaviour: The Role ofValues, Situational Factors and Goals’, Journal of Environmental Psychology,Volume 38, pp. 104–115.

Tan, M. (2014), Bad to worse: ranking 74 Chinese cities by air pollution.Greenpeace East Asia, http://www.greenpeace.org/eastasia/news/blog/bad-to-worse-ranking-74-chinese-cities-by-air/blog/48181/ (accessed 6December 2014).

Terpstra, T., Lindell, M.K. and Gutteling, J.M. (2009), ‘Does Communicating(Flood) Risk Affect (Flood) Risk Perceptions? Results of a Quasi-Experimental Study’, Risk Analysis, Volume 29, Number 8, pp. 1141–1155.

Terpstra, T. and Lindell, M.K. (2013), ‘Citizens’ Perceptions of Flood HazardAdjustments an Application of the Protective Action Decision Model’,Environment and Behavior, Volume 45, Number 8, pp. 993–1018.

Tobler, C., Visschers, V.H. and Siegrist, M. (2012), ‘Addressing ClimateChange: Determinants of Consumers’ Willingness to Act and to SupportPolicy Measures’, Journal of Environmental Psychology, Volume 32, Number 3,pp. 197–207.

Trainor, J.E., Murray-Tuite, P., Edara, P., Fallah-Fini, S. and Triantis, K. (2012),‘Interdisciplinary Approach to Evacuation Modeling’, Natural HazardsReview, Volume 14, Number 3, pp. 151–162.

Trainor, J.E., Nagele, D., Philips, B. and Scott, B. (2015), ‘Tornadoes, socialscience, and the false alarm effect’, Weather, Climate, and Society, Volume 7,Number 4, pp. 333–352.

Trumbo, C.W. and McComas, K.A. (2003), ‘The Function of Credibility inInformation Processing for Risk Perception’, Risk Analysis, Volume 23,Number 2, pp. 343–353.

Verroen, S., Gutteling, J.M. and Vries, P.W. (2013), ‘Enhancing Self-ProtectiveBehavior: Efficacy Beliefs and Peer Feedback in Risk Communication’, RiskAnalysis, Volume 33, Number 7, pp. 1252–1264.

Wachinger, G., Renn, O., Begg, C. and Kuhlicke, C. (2013), ‘The RiskPerception Paradox – Implications for Governance and Communication ofNatural Hazards’, Risk Analysis, Volume 33, Number 6, pp. 1049–1065.

Wei, J., Xu, J. and Zhao, D. (2015a), ‘Public Engagement with Firms on SocialMedia in China’, Journal of Information Science, Volume 41, Number 5, pp.624–639.

Wei, J., Zhao, M., Wang, F., Cheng, P. and Zhao, D. (2015b), ‘An EmpiricalStudy of the Volkswagen Crisis in China: Customers’ InformationProcessing and Behavioral Intentions’, Risk Analysis, Volume 36, Number 1,pp. 114–129.

Wei, J., Zhao, M., Wang, F. and Zhao, D. (2016), ‘The Effects of Firm Actionson Customers’ Responses to Product Recall Crises: Analyzing anAutomobile Recall in China’, Journal of Risk Research, Volume 19, Number4, pp. 425–443.

Whitmarsh, L. (2009), ‘Behavioural Responses to Climate Change:Asymmetry of Intentions and Impacts’, Journal of Environmental Psychology,Volume 29, Number 1, pp. 13–23.

Wu, H.C., Lindell, M.K. and Prater, C.S. (2015), ‘Process Tracing Analysis ofHurricane Information Displays’, Risk Analysis, Volume 35, Number 12, pp.2202–2220.

Xinhua, Qian. (2014), Smog and fog hit east, north China, 2013, http://www.china.org.cn/environment/2013-12/06/content_30814209.htm,(accessed 6 December 2014).

Zaleskiewicz, T., Piskorz, Z. and Borkowska, A. (2002), ‘Fear or Money?Decisions on Insuring Oneself Against Flood’, Risk, Decision and Policy,Volume 7, Number 03, pp. 221–233.

Zhang, Y. (2014), Severe winter smog shrouds eastern China, http://www.globaltimes.cn/content/829983.shtml (accessed 6 December 2014).

Zhang, A., Zhong, L., Xu, Y., Wang, H. and Dang, L. (2015), ‘Tourists’Perception of Haze Pollution and the Potential Impacts on Travel:Reshaping the Features of Tourism Seasonality in Beijing, China’,Sustainability, Volume 7, Number 3, pp. 2397–2414.

Zhu, W., Wei, J. and Zhao, D. (2016), ‘Anti-Nuclear Behavioral Intentions:The Role of Perceived Knowledge, Information Processing, and RiskPerception’, Energy Policy, Volume 88, pp. 168–177.

Journal of Contingencies and Crisis ManagementVolume 25 Number 4 December 2017© 2017 John Wiley & Sons Ltd

Protective Behaviors Towards City Smog in China 255