Journal Pre-proof

Risk perception and its impact on vegetable consumption: A case study from Hanoi,
Vietnam

Thanh Mai Ha, Shamim Shakur, Kim Hang Pham Do

PII: S0959-6526(20)32838-9
DOI: https://doi.org/10.1016/j.jclepro.2020.122793
Reference: JCLP 122793

To appear in: Journal of Cleaner Production

Received Date: 22 October 2019

Revised Date: 9 June 2020
Accepted Date: 10 June 2020

Please cite this article as: Ha TM, Shakur S, Pham Do KH, Risk perception and its impact on vegetable
consumption: A case study from Hanoi, Vietnam, Journal of Cleaner Production, https://doi.org/10.1016/
j.jclepro.2020.122793.

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.

© 2020 Elsevier Ltd. All rights reserved.

CRediT author statement
Thanh Mai Ha: Data collection, Data analysis, Writing- Original draft preparation. Shamim Shakur:
Supervision, Reviewing and Editing. Kim Hang Pham Do: Supervision and Reviewing.
 Risk perception and its impact on vegetable consumption: A
case study from Hanoi, Vietnam
Thanh Mai Haa,b,1, Shamim Shakura,2, Kim Hang Pham Doa,3
a
School of Economics and Finance, Massey University, Palmerston North 4442, New
Zealand
b
Faculty of Economics and Rural Development, Vietnam National University of Agriculture,
Trau Quy town, Hanoi, Vietnam
1
Corresponding author. Email t.ha@massey.ac.nz
2
Email: S.Shakur@massey.ac.nz
3
Email: K.H.PhamDo@massey.ac.nz
 1 Abstract
2 Vegetables form the mainstay of the Vietnamese diet, where food safety remains an on-going
3 concern. This paper investigates 1) the determinants of vegetable risk perception in the rural
4 and urban regions and 2) the impact of risk perception on vegetable consumption in Hanoi,
5 Vietnam. Using the data from a survey of 498 consumers, we found that food risk
6 information, perceived consequences of hazards, and institutional trust shaped risk perception
7 in both regions. Respondent’s age and education determined their risk evaluation in the rural
8 region only. Personal experience with vegetable poisoning, self-provisioning of vegetables,
9 and perceived control over hazards only influenced risk assessment in the urban region. 88%
10 of surveyed consumers avoided consuming vegetable varieties that were perceived unsafe,
11 switching instead to their safer alternatives. 33.5% of them ate fewer vegetables than before.
12 A heightened risk perception led to a reduction in vegetable consumption. Better
13 management of food safety and better risk communication toward attenuating consumer
14 anxiety about vegetable safety is essential to improve both producer and consumer welfare.
15
16 Keywords: food safety, risk perception, food consumption, vegetables, Vietnam
17
18
19 1. INTRODUCTION
20
21 Urban expansion poses challenges to ensuring food safety risks, particularly in perishable
22 categories such as meat, fish, and vegetables in Vietnam. The distance between primary food
23 producers and final consumers has lengthened, leading to an “information asymmetry” and
24 increased consumer uncertainty about food quality. The loss of farmland, farmer’s poor
25 knowledge, and government failure in managing the distribution of agrochemical inputs led
26 to the overuse of pesticides in vegetable production (Hoi et al., 2016). To foster sustainable
27 food production, Vietnam Ministry of Agriculture and Rural Development has introduced
28 “Safe Vegetable” program since 1995 and VietGAP (Vietnam Good Agricultural Practices)
29 in 2009 (Mergenthaler et al., 2009). Farmers in the “Safe Vegetables” program are provided
30 training on the proper use of fertilizers and pesticides. Producer groups who wish to obtain
31 VietGAP certificate need to record chemical applications and conduct internal inspections to
32 ensure food safety standards (Wang et al., 2014). Though the demand for high-quality and
33 safe vegetables is growing, consumer distrust in food labels and inadequate food safety
34 governance remain key barriers to market growth (Wang et al., 2014).
35
36 Vegetables are the mainstay of the Vietnamese diet. The possibility of chemical
37 contamination in vegetables has heightened food safety concerns among Vietnamese food
38 shoppers. In an earlier study by Figuié et al. (2004) on a sample of 200 Hanoi consumers,
39 vegetables were ranked as the most dangerous food, followed by meat, fruit, and then fish. 15
40 years later, Ha et al. (2019) surveyed about 500 Hanoi consumers, asking them to evaluate
41 the risk of 6 perishable food products. The respondents again rated vegetables as the riskiest
42 food. The fear of potential risks presented in vegetables remains persistent over a long period
43 even though many social, economic changes, and government efforts in addressing food
44 safety issues have taken place in Vietnam. Food anxieties can lead to the erosion of consumer
45 trust in vegetable products, resulting in economic losses for the sector. Addressing deep-
46 rooted food concerns remains a challenge for policymakers. Risk perception is
47 a subjective judgment that people make about the severity of risk (Slovic et al., 1982). An
48 understanding of factors that contribute to the heightened risk perception is essential in
49 developing efficient countermeasures to address food safety concerns.

1
50 Though vegetable eating is understood to bring multiple health benefits (Herbert et al., 2010),
51 consumers in many Asian developing countries like Vietnam and China are worried about
52 their safety aspects. The biggest concern is about vegetable contamination, especially
53 pesticide residues (Cheng et al., 2016, Wertheim-Heck et al., 2014). Despite this, research
54 that focuses on the determinants of vegetable risk perception in developing countries is
55 limited. This paper will address this gap.
56
57 Over 60% of Vietnam’s population lives in rural areas. Rural consumers differ from their
58 urban counterparts in economic activities, social linkages (Durkheim, 1933), and access to
59 healthy food (Larsen and Gilliland, 2008). Thus, factors affecting risk perception might be
60 different between these two consumer segments. An insight into such differences will
61 enhance the effectiveness of risk communication, which is suitable for each segment.
62 Nevertheless, many risk perception studies have ignored rural-urban dissimilarities in food
63 safety risk perception (Omari et al., 2018). Some studies compared risk perception between
64 the two regions by treating the residential location as a dummy predictor of risk perception
65 (Liu and Ma, 2016, McEachern and Warnaby, 2008). Little is known whether the
66 determinants of risk perception differ between regions. This paper compares the underlying
67 drivers of vegetable risk perception among regions for a comprehensive understanding of this
68 rural-urban divide.
69
70 A few incidents are known to shape negative feelings, risk perception, and risk-related
71 decision making (Burns et al., 2012). Schroeder et al. (2007) reported that BSE crises
72 heightened risk perception of beef, leading to their consumption in the United States and
73 Europe. In Vietnam, the prevalence of food safety incidents, along with increased media
74 attention contributes to food scares (Nguyen-Viet et al., 2017). Particularly for vegetables,
75 food safety worries prompt consumers to take precautions such as growing own vegetables
76 (Van Hoi et al., 2009) or purchasing conventional vegetables from trustworthy vendors
77 (Wertheim-Heck et al., 2014). Since vegetable safety concern is high in Vietnam, similar
78 behavior is likely to occur. However, the impact of risk perception on food consumption is
79 not fully researched in Vietnam and other developing countries. This paper investigates food
80 consumption as a consequence of heightened risk perception in a developing country context.
81
82 The objectives of this paper are twofold. First, we investigate and compare the determinants
83 of risk perception of vegetables between the rural and urban regions. Second, we analyse the
84 impact of risk perception on self-reported vegetable consumption reduction. The study
85 explores the similarities as well as disparities in influencing factors of risk perception
86 between rural and urban settings. The findings will enable decision-makers to design
87 effective risk communication strategies that are specific to each region. The information
88 about changes in food consumption due to food safety concerns will assist the development
89 of food safety policies in Vietnam and other developing countries that endure consumer
90 distrust in food.
91
92
93 2. CONCEPTUAL FRAMEWORK
94 We adopt the definition of Schroeder et al. (2007), where risk perception of vegetables is
95 defined as consumer evaluation of their health risk from consuming the product. A
96 conceptual framework in Figure 1 shows various psychological and socioeconomic factors
97 influencing risk perception of vegetables. Note that these factors might differ between the
98 rural and urban regions. A high risk perceived will lead to changes in vegetable consumption
99 patterns.
2
 Risk information Institutional trust
acquisition

104
Rural versus urban

105
106
Food safety risk Vegetable consumption
perception of vegetables changes (avoiding,
switching, reducing)

112
113
Risk characteristics of Social economic factors114
hazards -Home-grown vegetables115
-Perceived control -Demographic 116
-Perceived consequence characteristics 117
-Perceived knowledge -Vegetable poisoning 118
experience 119
120
121 Figure 1: Conceptual framework of factors influencing vegetable risk perception that
122 regulates vegetable consumption
123
124
125 2.1. Determinants of risk perception of vegetables
126
127 Studies have examined the rural-urban divide in risk perception, information access, and
128 trust. While a survey in China reported that urban people hold a higher food risk perception
129 than their rural counterparts (Liu and Ma, 2016), research from Scotland and the UK provide
130 contradictory findings (Hall and Moran, 2006, McEachern and Warnaby, 2008). Rural and
131 urban regions also differ in their search for health information. Rural people were less likely
132 to use the internet for health purposes due to lower educational level and income (Hale et al.,
133 2010). There is evidence that trust in institutions is lower in the urban region. Shi (2001)
134 found that more media access eroded urban people’s trust in the government. Since rural and
135 urban consumers are dissimilar in their risk judgment and many other aspects, we expect that
136 factors influencing risk perception might differ between these two consumer groups.
137
138 Institutional trust refers to trust in institutions that are responsible for managing food safety.
139 According to Siegrist (2000), trust plays an essential role in the absence of knowledge. A
140 negative association between lack of trust in institutions and food safety risk perception has
141 been empirically established by Knight and Warland (2005) and Akbari et al. (2019).
142 Interestingly, Knight and Warland (2005) reported that trust in the food system was a good
143 predictor of risk perception, even when people believe they have knowledge and control over
144 risks. Based on their evidence, we hypothesize a negative relationship between institutional
145 trust and risk perception.
146
147 The psychometric paradigm developed by Slovic (1992) is one of the major theories on risk
148 perception. This psychological-based theory argues that individuals’ risk interpretation of a

3
149 hazard is formed by risk characteristics - the extent the hazard is perceived to be known,
150 controllable, and severe. Different hazards might contaminate vegetables. Chemical hazards
151 are evaluated by consumers to be more dangerous than bacterial hazards because of the belief
152 that these hazards are more severe, less familiar, and less controllable (Kher et al., 2013). It
153 is expected that risk perception of vegetables is shaped by risk characteristics of hazards
154 embodied in the product. Perceived control and knowledge of hazards will reduce risk
155 perception while the perceived consequence of hazards will increase risk perception of
156 vegetables.
157
158 Social amplification of risk (Kasperson et al., 1988) provides an understanding of the role of
159 risk information in shaping risk perception. This theory also explains how a small risk, as
160 assessed by experts, results in public concerns with profound impacts. When an incident
161 occurs, information about risk is collected, processed, and interpreted by individuals. Risk is
162 then amplified through the interaction between individuals and social amplification stations
163 (e.g., scientists, the media, and social networks). Having experienced a food incident,
164 confidence in food safety will drop dramatically, particularly when new risk information
165 spreads. Once this information is captured and exploited by mass media, risk perception of
166 food escalates, causing unwarranted panics (Verbeke, 2005). We expect that food risk
167 information acquisition is positively associated with the risk perceived of vegetables.
168
169 According to the cultural theory developed by Douglas and Wildavsky (1983), risk
170 perception has a social or cultural construct. Individuals vary in their risk perception, and
171 collectively there is disagreement about the most dangerous risk to society. Dosman et al.
172 (2001), in their review, revealed the significant influence of demographic variables such as
173 age, education, income, and the presence of children on risk perception. Homegrown
174 vegetables are believed very safe. Access to homegrown vegetables represents better control
175 over vegetable safety. This, in turn, may reduce the perceived risk of vegetables. Direct
176 experience from food poisoning in the past might evoke memory about the event and increase
177 risk perception (Green et al., 2003).
178
179 2.2. The influence of risk perception on food consumption
180
181 According to the psychological approach, the more a person dreads an activity, the higher
182 risk perception and risk-reducing willingness she/he holds (Slovic et al., 1982). We assume
183 that consumers, in general, are risk-averse, who strive to lower the probability of negative
184 impacts from vegetable poisonings. In response to perceived risk, risk-averse consumers will
185 modify their consumption decision and develop risk-reducing strategies. Consumers tend to
186 avoid eating particular food categories if they feel a significant threat from it. The avoiding
187 strategy will be used if consumers do not like the food, or they can find substitutes (Järvelä et
188 al., 2006). An avoidance strategy when coupled with a substitution strategy may keep total
189 vegetable consumption stable. The total consumption would invariably decline in the absence
190 of safer food alternatives.
191
192 The association between risk perception and food consumption outcomes has been well
193 documented. Yeung and Morris (2006) revealed the negative relationship between risk
194 perception and the likelihood of purchasing chicken meat in Britain. Similarly, Pennings et al.
195 (2002) found that heightened risk perception of mad cow disease across the US and European
196 countries increased the probability that consumers reduced beef consumption. Schroeder et
197 al. (2007) in their survey, also on the US and Europe, reported that risk perception of mad
198 cow disease determined not only the probability but also the percentage of self-reported

4
199 reduction in beef consumption. We hypothesize a positive correlation between risk perception
200 and vegetable consumption reduction. In this study, the reduction in vegetable consumption
201 represents the volume of fresh vegetables that consumers have cut down within a period due
202 to concerns about vegetable safety.
203
204 3. DATA AND METHOD
205
206 3.1 Study area and consumer survey
207 We conducted a consumer survey in Hanoi, Vietnam, from February to April 2017. A high
208 level of concern about vegetable safety of Hanoi residents offered an opportunity to study
209 how such worry influenced vegetable consumption. A diverse population of more than 8
210 million people living in urban as well as rural districts of Hanoi offered a perfect setting for a
211 rural-urban comparison.
212
213 We drew a sample from 4 urban and 3 rural from a total of 30 districts of Hanoi representing
214 varying levels of economic development. The Vietnamese Government classifies rural and
215 urban districts. We applied quota sampling with a predetermined number of respondents
216 (from 70 to 80) from each district. With support from community leaders, oral invitations to
217 take part in the survey were sent to primary food shoppers who are at least 18 years old. In
218 each district, we selected from 70 to 80 agreed participants with varied incomes. The face-to-
219 face survey, using structured questionnaires was conducted at respondent’s homes. The
220 sample consists of 498 respondents, of which 230 are from the rural and 268 from the urban
221 region. This sample is not representative of Vietnam’s population, as it had a high percentage
222 of female and highly educated respondents (Table 1).
223
224 3.2. Variable measurement
225
226 3.2.1. Variables used to analyze the determinants of risk perception
227
228 We measured variables relating to trust and perception on a 10-point Likert scale where a
229 higher scale reflects a higher level of trust and perception (Table 1). Our pilot survey
230 demonstrated that a 10-point scale reminded respondents of the academic grading system in
231 Vietnam with which they are familiar.
232
233 Risk perception of vegetable (PV), the dependent variable, was measured by the question
234 item adapted from Schroeder et al. (2007): “To what extent do you think that you are exposed
235 to the health risk from eating vegetables you have consumed.” For a better interpretation of
236 the results from ordered logit regressions, this variable was later transformed from an initial
237 10-point into a 5-point scale representing 5 ordered categories of risk perceived (very low,
238 low, moderate, high, and very high). Such transformation, according to Dawes (2008), does
239 not cause significant changes in data characteristics.
240
241 We proposed 5 constructs as potential predictors of risk perception. 5 items formed trust
242 construct. Trust in food traders in the wet market was the lowest. Generally, urban people
243 placed a lower level of institutional trust than rural residents (Table 1).
244
245 “Perceived Knowledge” conveys knowledge level respondents believed to have acquired
246 across various hazards. “Perceived Control” reflects consumers’ view about the extent they
247 were able to reduce particular hazards through washing and cooking vegetables. “Perceived
248 Consequence” refers to the evaluation of adverse health impacts from hazards. These

5
249 constructs were measured by recognized scales of Yeung and Morris (2006) to capture risk
250 characteristics of hazards.
251 “Information” measures the frequency of information acquisitions about food incidents via
252 mass media (TV), social media (Facebook), and personal sources (relatives and friends).
253 These are the top three preferred information channels of consumers to access food safety
254 news (Rutsaert et al., 2013). Responses ranged from 1 to 5, where 1 means “never” and 5
255 means “always” - having heard nearly every day.
256
257 Vegetable poisoning is defined as having acute food poisoning symptoms after eating
258 vegetables. The rest are dummy variables presenting socioeconomic factors that can predict
259 risk perception.
260
261 Table 1: Variable definition and statistics
Variables/ Variable definition Scale Whole Rural Urban
constructs sample
(n=498) (n=230) (n=268)
Mean (SD) Mean (SD) Mean (SD)
PV Risk perception of vegetables 1-5 3.80 (0.96) 3.60* (1.01) 3.90* (0.90)
Trust in Local government 1-10 4.11 (2.95) 5.28* (3.15) 3.11* (2.34)
Central government 1-10 3.56 (2.54) 4.32* (2.75) 2.91* (2.14)
Supermarkets 1-10 4.46 (2.44) 4.50 (2.66) 4.43 (2.25)
Farmers 1-10 2.77 (2.18) 3.16* (2.44) 2.44* (1.87)
Food traders at wet markets 1-10 2.42 (1.85) 2.77* (2.13) 2.13* (1.52)
Perceived Pesticides 1-10 5.17 (2.52) 5.09 (2.68) 5.23 (2.38)
knowledge of Bacteria 1-10 4.90 (2.33) 4.76* (2.38) 5.02* (2.28)
Heavy metal 1-10 4.19 (2.38) 3.97* (2.40) 4.39* (2.36)
GMO vegetables 1-10 3.38 (2.41) 3.13* (2.41) 3.59* (2.40)
Perceived Pesticide residue 1-10 3.45 (2.19) 3.75* (2.31) 3.20* (2.05)
control Bacteria 1-10 4.93 (2.56) 4.73 (2.48) 5.11 (2.62)
over Heavy metal 1-10 2.86 (2.01) 3.0*6 (2.06) 2.70* (1.96)
GMO vegetables 1-10 2.70 (2.10) 3.09* (2.27) 2.36* (1.88)
Perceived Pesticide residues 1-10 8.20 (1.98) 8.13 (2.05) 8.27 (1.91)
consequence Bacteria 1-10 7.28 (2.19) 7.12 (2.33) 7.42 (2.06)
of Heavy metal 1-10 7.90 (2.09) 7.58* (2.29) 8.19* (1.86)
GMO vegetables 1-10 7.41 (2.51) 7.01* (2.73) 7.76* (2.25)
Information Food incidents from TV 1-5 3.92 (0.82) 3.81* (0.80) 4.01* (0.84)
about Food incidents from social media 1-5 3.34 (1.27) 2.83* (1.35) 3.78* (1.01)
Food incidents from 1-5 3.68 (0.93) 3.54* (0.91) 3.81* (0.93)
relatives/friends
VegGrow =1 if the family grow vegetables 0-1 0.60 (0.49) 0.85* (0.36) 0.38* (0.48)
Veg =1 if have been poisoned by 0-1 0.31 (0.46) 0.31 (0.46) 0.30 (0.46)
Poisoning vegetables
Elderly =1 if >= 60 years old 0-1 0.37 (0.38) 0.48* (0.50) 0.28* (0.44)
Male =1 if male 0-1 0.13 (0.33) 0.12 (0.33) 0.13 (0.33)
Rich =1 if individual monthly income 0-1 0.25 (0.43) 0.09* (0.28) 0.40* (0.49)
>=15 mil VND
Children =1 if there was at least 01 child in 0-1 0.76 (0.43) 0.69* (0.46) 0.81* (0.39)
the family
University =1 if have a university degree 0-1 0.51 (0.50) 0.25* (0.43) 0.74* (0.44)
262 Note: 23000 VND=1 USD. * The mean scores in the same row are statistically different at 5%, using
263 independent sample T-test.
264

6
265 3.2.2. Variables used to analyze changes in self-reported vegetable consumption and the
266 impact of risk perception on vegetable consumption
267
268 To capture changes in food consumption caused by risk perception, our questionnaire began
269 with a filter item, asking whether respondents worry about vegetable safety. Those having a
270 positive response then answered the question: “Due to the concern about food safety, have
271 you avoided eating any particular vegetables in the last two years? “If yes, please list their
272 names”. Food consumption patterns may vary over time for other reasons like price
273 fluctuations, change in marital status or parenting. The filter above eliminates response
274 errors, as it captures only the consumption changes due to food safety reasons.
275
276 Two items adapted from Schroeder et al. (2007) were used to reveal 1) whether the
277 respondent reduced vegetable consumption during the last two years due to the concern about
278 food safety, and 2) the estimated percentage reduction in vegetable consumption by volume.
279 The chosen period witnessed some major scandals on vegetable safety in 2015 that can
280 potentially impact on vegetable consumption. These two scandals are discussed later in
281 section 5. The variables used in the analysis are shown in Table 2. The first four were used to
282 examine changes in vegetable consumption. The remaining two were to analyze the impact of
283 risk perception on the consumption.
284
285 Table 2: Variables used to analyze changes in vegetable consumption
Aspects Variables
Changes in % of respondents reduced vegetable consumption volume
vegetable consumption % of vegetable consumption volume to be
reduced/respondent
% of respondents avoided eating at least one vegetable
Number of vegetables to be avoided/respondent
Impact of risk perception Risk perception
on consumption Level of consumption reduction
286 Note: There are 4 levels of consumption reduction. Level 1 denotes a decrease ≤ 20% of
287 totalvegetable consumption, level 2 (from 20-39%), level 3 (from 40-59%), and level 4 (≥
288 60%).
289
290 3.3. Data analysis
291
292 3.3.1 Analyzing the determinants of risk perception
293
294 The determinants of risk perception were identified in 2 steps: 1) principal component
295 analysis (PCA) and 2) ordered logit regression. PCA helps reduce the large and correlated
296 dataset into a fewer number of uncorrelated principal components while retaining ample
297 information of the dataset (Jolliffe, 2002). Since 20 independent variables measuring the
298 constructs are highly correlated, they were subjected to PCA, using VARIMAX rotation. We
299 performed separate PCAs for the rural and urban subsamples, but the results are not different
300 across subsamples. The components retained in the rural dataset are similar to those in the
301 urban dataset. We, therefore, performed PCA on the pooled data. Components retained from
302 PCA can be interpreted as potential determinants of risk perception. We employed two
303 criteria to select components: the eigenvalue of the component being greater than one, and
304 the total variance explained by all chosen components is at least 60%.

7
305
306 The components extracted from PCA, together with socioeconomic variables, were regressed
307 separately over regions. This compares how the rural and urban regions differ in the
308 determinations of risk perception. As risk perception of vegetables has 5 ordinal outcomes
309 (very low, low, moderate, high, or very high), ordered logit regression, which is suitable for
310 the model with an ordinal dependent variable was employed.
311
312 We assessed two assumptions of ordered logit regression: multicollinearity and proportional
313 odds. Correlation coefficients of all pairs of independent variables were smaller than 0.6,
314 demonstrating that multicollinearity was not a problem. The approximate likelihood-ratio test
315 was significant for the rural (χ2(39) = 66.69, p = 0.003) as well as the urban sample (χ2(35) =
316 54.85, p = 0.017). The second assumption was met, suggesting that there is the same set of
317 coefficients across different outcomes of risk perception across rural and urban models.
318
319 3.3.2. Analyzing the influence of risk perception on vegetable consumption
320
321 We used the Kruskal-Wallis test (Kruskal and Wallis, 1952) to examine the impact of risk
322 perception on vegetable consumption reduction that does not follow a normal distribution
323 (Table 2). This is a nonparametric statistical test for non-normalised data. The test helps
324 identify whether vegetable consumption reduction levels differ among 5 groups of
325 respondents, based on 5 risk perception (ordinal) outcomes. All observations in the whole
326 sample were ranked ascendingly for “Consumption reduction level”. We then calculated the
327 mean rank for each group where a higher mean rank implied a higher level of consumption
328 reduction. The Dunn-Bonferroni post hoc test was then conducted to examine whether the
329 rank means are statistically significantly different between each pair within 5 groups.
330
331 4. RESULTS
332
333 4.1. Determinants of risk perception of vegetables
334 Table 3 shows the results of PCA. The Kaiser-Meyer-Olkin (KMO) that measures sampling
335 adequacy was at an acceptable level (0.72). Cronbach's alphas of all the components were
336 higher than 0.6, suggesting adequate construct reliability (Tavakol and Dennick, 2011). 5
337 components were selected (Table 3). These retained components were able to explain about
338 65% of the total variance of the data set.
339
340 Table 3: Results of Principal Components Analysis
n = 498
Items Components
1 2 3 4 5
Trust Perceived Perceived Perceived Inform
knowledge consequence control
Trust in central government 0.83
Trust in local government 0.81
Trust in supermarkets 0.78
Trust in retailers at wet markets 0.71
Trust in farmers 0.63
Perceived knowledge of heavy metals 0.89
Perceived knowledge of pesticides 0.85
Perceived knowledge of bacteria 0.83

8
 Perceived knowledge of GMO 0.71
Perceived consequence of heavy metal 0.88
Perceived consequences of pesticides 0.73
Perceived consequences of bacteria 0.87
Perceived consequences of GMO 0.77
Perceived control over heavy metal 0.89
Perceived control over pesticide 0.77
Perceived control over bacteria 0.79
Perceived control over GMO 0.72
Food incidence heard from relatives 0.75
Food incidence heard from social media .78
Food incidence heard from TV 0.72
% of total variance explained 19.95 17.78 12.81 8.46 6.84
341 Note: The Cronbach’s alpha of the 1st, 2nd, 3rd, 4th, 5th component are 0.81, 0.85, 0.84, 0.75, 0.65, respectively
342 Source: Authors’ calculation
343
344
345 Table 4 provides the results of ordered logit regression by region. For brevity, we only
346 reported marginal effects of the highest category of dependent variables - “very high risk.”
347
348
349 Table 4: Ordered logit regression results
Variables Rural Urban
Coefficient Marginal Coefficient Marginal
effect effect
(SE) (SE) (SE) (SE)
Male -0.442 -0.063 -0.255 -0.047
(0.406) (0.058) (0.358) (0.066)
Elderly 0.668* 0.096 -0.056 -0.010
(0.354) (0.050) (0.496) (0.092)
University 0.692** 0.099 -0.025 -0.004
(0.323) (0.045) (0.283) (0.052)
Children 0.025 0.003 -0.252 -0.047
(0.271) (0.038) (0.303) (0.056)
Rich 0.035 0.005 0.193 -0.036
(0.452) (0.065) (0.253) (0.047)
Veg Poisoning 0.163 0.023 0.570*** 0.106
(0.293) (0.042) (0.260) (0.047)
Veg Grow -0.430 -0.061 -0.408** -0.076
(0.352) (0.050) (0.206) (0.045)
Perceived Knowledge 0.117 0.017 0.201 0.037
(0.127) (0.018) (0.125) (0.023)
Perceived Consequence 0.287** 0.041 0.591*** 0.110
(0.121) (0.017) (0.138) (0.024)
Perceived Control -0.088 -0.013 -0.472*** -0.088
(0.128) (0.018) (0.126) (0.022)
Information 0.512*** 0.073 0.284** 0.053
(0.143) (0.021) (0.140) (0.025)
Trust -0.293** -0.042 -0.265* -0.049
(0.124) (0.018) (0.143) (0.026)
Log-likelihood ratio (p=0.000) 36.84 51.68
9
 Count R2 40.00 49.60
350
351 Note: Numbers in brackets are standard errors;***, **,*: significant at 1%, 5%, 10% level,
352 respectively; SE denotes standard error.
353
354
355 No demographic variable exerted a significant effect on the urban region, while only age and
356 education affected risk perception in the rural region. Urban residents who experienced
357 vegetable poisoning was 10.6% more likely to report a very high-risk of vegetables than
358 those who have not. Self-provisioning of vegetables and perceived control over hazards
359 reduced the chance that urban respondents reported “very high risk”.
360
361 In both regions, the more frequently respondents heard about food safety events, the higher
362 was the probability they cited a very high risk. Trust caused a negative effect, while the
363 perceived consequence of hazards generated a positive impact on risk perception.
364
365 4.2. Risk perception and its impact on vegetable consumption
366 All surveyed consumers stated that they were worried about vegetable safety. Risk perception
367 was considerably high for the whole sample (mean = 3.8 (out of 5) ± 0.96). The two-sample
368 T-test shows a statistically significant difference between the two regions regarding risk
369 perception (3.9 ±0.9 for the urban and 3.6 ± 1.00 for the rural, p = 0.00) (Table 1).
370
371 About one-third of respondents decreased the volume of vegetable intake during the last 2
372 years because of food safety concerns (Table 5). The self-reported reduction was about 8.5%
373 per respondent. 88% of consumers surveyed excluded at least one vegetable variety from
374 their diet. Among the top 10 frequent-cited vegetables, 7 of them belong to a leafy category
375 (pak choy, choy sum, cabbage, broccoli, morning grow, watercress, lettuce). Regarding the
376 type of consumption, 4 of them are often eaten as a salad (Thai brinjal, cucumber, bean
377 sprout, and lettuce) while the remaining 6 are usually cooked.
378
379 Table 5: Changes in vegetable consumption due to food safety concerns
Indicators Whole Rural Urban
sample
% of respondents reported vegetable consumption reduction 33.50 34.34 33.84
% of vegetable consumption has been reduced/respondent 8.47 8.46 8.54
% of respondents avoided eating at least one vegetable 88.15 87.50 88.79
Number of vegetables that were avoided eating/respondent 2.23 2.16 2.29
Top 10 vegetables that were frequently avoided eating pak choy, choy sum,
cabbage, broccoli,
morning grow,
watercress, Thai brinjal,
cucumber, bean sprout,
and lettuce
380 Note: The scientific name of pak choy, choy sum, morning grow, and Thai brinjal is
381 Brassica rapa L. var. chinensis, Brassica rapa var. parachinensis, Ipomoea aquatic Forsk,
382 and Solanum macrocarpon L., respectively.
383
384 Though risk perception of vegetables was higher in the urban region, the self-reported
385 decrease in vegetable consumption tends to be similar between the rural and urban regions.
386 The values of the first four indicators in Table 5 are not much different across the two

10
387 regions. The chi-square test (for the first and the third indicator) and two sample T-test (for
388 the second and the fourth indicator) between rural and urban groups were insignificant,
389 indicating that the two regions shared the same trend in reduction of vegetable consumption.
390
391 Table 6 shows a statistically significant difference in vegetable consumption reduction
392 between different levels of risk perceived from vegetables (χ2(4) = 45.135, P = 0.000). This
393 suggests that at least one group had the mean rank of vegetable consumption reduction
394 different from other groups. This mean rank increased with risk perception levels, suggesting
395 a positive association between risk perception and consumption reduction. The post hoc test
396 further revealed that the mean rank of vegetable consumption reduction differed significantly
397 across risk perception outcomes either at 5% or 10% level (Table 7). All of these results
398 confirm that a higher risk perceived led to a higher level of vegetable consumption reduction.
399
400 Table 6: Risk perception level and vegetable consumption reduction
n = 498
Risk perception outcome N Mean rank of consumption reduction
level
1 10 178.00
2 34 198.78
3 128 233.01
4 197 237.05
5 129 303.79
χ2(4) = 45.135, P =
0.000
401
402 Note: Reduction in vegetables consumption is in scale 1-4 with 1 (reduce less than 20%), 2 (reduce
403 20% -39 %), 3(reduce 40% - 59%), 4(reduce more than 60%); risk perception for vegetables was in a
404 1-5 scale with 1(very low) and 5(very high). The Kruskal-Wallis test calculates the mean rank of
405 consumption reduction level.
406
407 Table 7: Results of post hoc test
Compare the mean rank of consumption reduction χ2(1) P
between risk perception outcome
1 and 2 0.93 0.34
1 and 3 2.88 0.09*
1 and 4 3.04 0.08*
1 and 5 8.50 0.04**
2 and 3 3.30 0.07*
2 and 4 4.00 0.04**
2 and 5 18.41 0.00**
3 and 4 0.10 0.74
3 and 5 21.95 0.00**
408 Note**, *: significant at 5% and 10% level, respectively, using Kruskal- Wallis test
409
410
411 5. DISCUSSION
412
413 5.1. Factors affecting risk perception

11
414
415 We found notable differences in the determinants of vegetable risk perception between the
416 rural and urban regions. These determinants are demographic variables, personal experience
417 with vegetable poisonings, self-provisioning of vegetables, and perceived control over
418 hazards.
419
420 While age and education shaped vegetable risk perception in the rural area, no demographic
421 variable influenced risk evaluation in the urban region. Unlike Lee et al. (2012), we found a
422 positive association between education and risk perception, but significant only for the rural
423 subsample. In the rural region, better education might be associated with a higher intention to
424 seek food safety (risk) information (Yang et al., 2020), leading to higher risk perception. In
425 the urban region, risk perception did not differ across different levels of education. More
426 research is required to retest the influence of education on risk perception in the urban setting.
427 For urban consumers, their personal experience with food poisoning increased their risk
428 perception, as direct exposure to risk events often enhances consumer’s memory and
429 imagination of hazard (Kasperson et al., 1988). The effect of “vegetable poisoning” in the
430 rural region was insignificant because vegetable poisonings were less severe, as reported by
431 rural residents. A possible reason is that eating out is less common in rural areas (Lachat et
432 al., 2011).
433
434 The significant effect of self-provisioning of vegetables in the urban area is explained by the
435 belief that homegrown vegetables are safe (Green et al., 2003). We found that despite land
436 scarcity, 38% of urban households were growing vegetables for family consumption. These
437 households undertook initiatives to grow pesticide-free vegetables on rooftops or strips of
438 adjacent public land. The use of chemical fertilizers was either absent or kept at a minimum.
439 Homegrown vegetables, in respondents’ eyes, were thus absolutely safe. Furthermore, due to
440 food safety concerns, not only the poor but also the rich are growing own vegetables in urban
441 districts. The mean monthly family expenditure was not significantly different between those
442 growing own vegetables (102 households) and those who are not (166 households) (11.3
443 million VND versus 11.5 million VND, p = 0.204, independent sample T-test). However, in
444 the rural region, risk perception was not related to homegrown vegetables. It might be due to
445 the lack of variance of this variable in the rural sample as 85 % of rural households grow
446 some vegetables.
447
448 Perceived control over hazards statistically decreased risk perception only in the urban
449 region, suggesting that improving urban consumers’ perceived control over hazards will help
450 reduce anxiety about vegetable safety. Education program on food hazards in the urban
451 region is necessary to enhance perceived control over food hazards.
452
453 Rural and urban regions shared three similar predictors of risk perception: food risk
454 information acquisition, trust, and perceived consequence of hazards. Risk perception in both
455 regions increased with information acquisition about food safety incidents. Kasperson et al.
456 (1988) indicated that media and social groups interact with each other to escalate a marginal
457 risk into a heightened risk. Such media effect is also found in Vietnam. In popular national
458 television programs like “Say No To Contaminated Food”, there is an unbalanced coverage
459 on negative news about food safety compared to good remedial news. Since bad news tend to
460 stay more with the viewers than good news (Slovic, 1993), frequent food safety scandals, and
461 the media effect led to food scares.
462

12
463 A negative relationship between risk perception and trust was found across the regions. Trust
464 in the food system acts as a coping mechanism that helps consumers reduce the complexity of
465 risk assessment and therefore eliminate their concerns (Knight and Warland, 2005). The
466 finding above suggests that reducing risk perception requires an improvement in institutional
467 trust.
468
469 In line with the psychometric paradigm, this paper found that the perceived consequence of
470 hazards accelerated risk perception across regions. However, other risk characteristics, for
471 example, perceived knowledge of hazards did not translate into risk perception of vegetables.
472 More empirical research on other food products is needed to retest the link between different
473 risk characteristics of hazards with risk perception of a particular food.
474
475 5.2.Risk perception of vegetables and their impact on vegetable consumption

476 Risk perception of vegetables was higher in the urban region. Perhaps, this is due to the
477 absence of homegrown vegetables in 62% of urban households, comparing to 15% of rural
478 families, as mentioned earlier. Furthermore, urban residents might have better access to food
479 safety information (Hale et al., 2010), which probably leads to a higher perceived risk.
480
481 Heightened risk perception made consumers selective in consumption by avoiding vegetable
482 species that are considered unsafe. When confronted with a high perceived risk from
483 vegetables, consumers tend to employ optimal risk reduction strategies by maximizing their
484 utility given available resources. Rather than engaging in a costly information search, 88% of
485 respondents chose an uncomplicated strategy of not consuming vegetables perceived as risky.
486 The percentage of consumers who avoided eating at least one vegetable variety was higher
487 than the proportion of consumers that reduced the total vegetable consumption (Table 5).
488 This result indicates that when postponing consumption of some vegetables, consumers
489 switched to alternatives that are considered safer.
490
491 Similar to a survey conducted on a small city of Vietnam by Wertheim-Heck et al. (2014),
492 our survey uncovered that leafy vegetables were deemed to be more unsafe compared to root
493 vegetables (carrot, potato) and fruits (hairy melon, guava bean, sponge luffa). Consumers
494 believed that root and fruit vegetables with removable skin could prevent these vegetables
495 from absorbing chemicals. Because of this perception, consumers may shift from leafy
496 vegetables to alternative categories. Risk-avoiding behavior has been investigated by Green
497 et al. (2003). They found that consumers avoided food categories that were considered to be
498 harmful, such as frozen, imported, ready to eat food or takeaway food. Our research
499 complements previous studies that depict a common characteristic of most of the consumers:
500 the risk-averse behavior in vegetable consumption.
501
502 Among the top ten riskiest vegetables listed in Table 5, the first seven were thought to be
503 contaminated by insecticides while the next two were with growth regulators that are
504 commonly used during vegetable production in Vietnam. For lettuce, the last item in Table 5,
505 consumers believed that contamination was caused by eating them raw, which, unfortunately,
506 is the usual way of consuming the product. In line with Van Hoi et al. (2009), this research
507 suggests that pesticide residues in vegetables are huge concerns to consumers in Vietnam, as
508 shown by a high mean score of perceived consequences of pesticide in Table 5. As a result,
509 one-third of respondents chose to eat fewer vegetables than before, while the remainder
510 decided to shift to safer alternatives. Such behavior can lead to less diverse vegetable intake.
511

13
512 An increase in risk perception led to a reduction in vegetable consumption. Earlier research
513 by Figuié and Fournier (2008) also found such consumer reaction in Vietnam. The authors
514 indicated that during the first outbreak of Avian Influenza Virus, H5N1 in January 2004, 74%
515 of consumers stopped eating chicken. Two years later, 6% of consumers still avoided the
516 product. Our study found that the average percentage reduction in vegetable consumption per
517 consumer was rather small at 8.5%. This reduction might be due to vegetable safety scandals
518 reported in mass media in the past. For example, in January 2015, some big supermarkets in
519 Hanoi were reported to trade conventional vegetables but claimed these vegetables “safe”
520 (Vietnam News, 2015). The incident publicized by various media outlets impacted negatively
521 on supermarkets’ image. In our survey conducted two years after this scandal, risk perception
522 of vegetables was still high, causing a marginal decrease in vegetable consumption. This
523 scandal and other food incidents might impact vegetable consumption. Our study suggests
524 that a decline in per capita vegetable consumption in Hanoi is a consequence of heightened
525 risk perception and poor food safety management. However, this finding should be
526 interpreted with caution, as this cross-sectional study does not allow for investigation of the
527 causal link among food safety scandals, risk perception, and food consumption over time.
528
529 6. CONCLUSIONS AND POLICY IMPLICATIONS

530 We found that some of the factors influencing vegetable risk perception were similar, while
531 others differed between rural and urban regions. Regional characteristics should be taken into
532 account in the development of food risk communication and food safety policy. Because age
533 shaped risk perception in the rural region, adequate food risk communication for the old is
534 essential. In the urban region, consumer education programs on food hazards would help
535 improve consumers’ perceived control over the hazards, thereby reducing their amplified risk
536 perception.
537
538 Inadequate risk information sharing across regions augmented risk perception of vegetables.
539 Prudent risk communication should be based on the factual evidence about the risk, not just
540 focusing on negative news but also educating consumers on remedial aspects of food safety.
541 Recent government initiatives like “Action month for food hygiene and safety” had a
542 temporary effect on consumer's knowledge. Long term approaches, for example, early
543 education of food safety can be a better solution. Since trust in government and food industry
544 is low, risk information should not be communicated by them but by the most trusted sources
545 like health professionals and environmental organizations.
546
547 Improving trust is essential for both regions, as trust determined risk perception. Consumers
548 have observed corruption and widespread rent-seeking behavior of government authorities
549 such as food inspectors (Van Hoi et al., 2009). A series of food scandals have severely
550 destroyed trust in the food industry. To restore trust, the government needs to demonstrate its
551 commitment to food safety. Trust in the food industry can be rebuilt by providing truthful
552 information about food products and complying with food safety regulations.
553
554 This paper highlights the importance of urban farming, as the self-provisioning of vegetables
555 becomes a norm in big cities in Vietnam. Hanoi has a current population of over 8.5 million,
556 with an annual growth rate of 3.5%. Ensuring food security and food safety for such a
557 growing population remains a challenging task. Pulliat (2015) found that Hanoi households
558 that engaged in urban agriculture were mainly low-income families. Self-provision of food, a
559 livelihood strategy for these families contributed to household food security. Our survey
560 revealed that due to food safety concerns, growing own vegetables was popular not only

14
561 among low-income, but also wealthy families in the urban region. We found that urban
562 farming, a casual household food production practice, can contribute to sustainable food
563 consumption in metropolitan cities by enhancing food safety and reducing food fears.
564 Environmentally-friendly urban agriculture, such as integrated rooftops should be
565 encouraged.
566
567 In response to a heightened risk perceived from vegetables, besides growing their vegetables,
568 consumers modified their vegetable consumption and employed several risk-reducing
569 strategies. The majority of consumers switched from perceived-unsafe vegetable categories
570 (leafy vegetables) to perceived-safer alternatives (root and fruit vegetables). Eating fewer
571 vegetables was another strategy of about one-third of consumers. Risk perception was
572 significantly related to vegetable consumption reduction.
573
574 Vegetable safety scandals reported in mass media in the past might have caused an impact on
575 risk perception that led to a reduction in vegetable consumption. This will have a noticeable
576 negative impact on the vegetable sector of Vietnam. Heightened risk perception also implies
577 some consequences for consumers. Reducing consumption and being selective in vegetable
578 choice might hinder consumers’ freedom in eating vegetables and the diversity in vegetable
579 intake. Having better control of vegetable safety and reducing heightened risk perception of
580 vegetables would help boost vegetable demand and enhance the sustainability of vegetable
581 production in Vietnam.
582
583 This paper has some limitations. Firstly, using a single item to measure risk perception might
584 be inadequate to capture the complexity of risk perception construct that is inherently
585 multidimensional. Secondly, since our sample is limited to Hanoi consumers only, the
586 research finding should not be generalized to the entire population of Vietnam. Thirdly, this
587 paper is based on cross-sectional data, which is not the best to establish the causal
588 relationship between risk perception and vegetable consumption reduction. Future studies
589 that use a multi-item measurement scale for risk perception and are based on a national
590 representative sample from panel data can address these limitations.
591
592
593 Declaration of interest: None
594
595 Acknowledgment
596
597 We appreciate the support from the staff at the Faculty of Economics and Rural
598 Development, Vietnam National University of Agriculture, Vietnam during the survey.
599
600 Funding
601 This research did not receive any specific grant from funding agencies in the public,
602 commercial, or not-for-profit sectors.
603
604 REFERENCES
605
606 Akbari, M., Ardekani, Z. F., Pino, G. & Maleksaeidi, H. 2019. An extended model of Theory
607 of Planned Behavior to investigate highly-educated iranian consumers’ intentions
608 towards consuming genetically modified foods. Journal of cleaner production, 227,
609 784-793.

15
610 Burns, W. J., Peters, E. & Slovic, P. 2012. Risk perception and the economic crisis: A
611 longitudinal study of the trajectory of perceived risk. Risk Analysis: An International
612 Journal, 32, 659-677.
613 Cheng, L., Jiang, S., Zhang, S., You, H., Zhang, J., Zhou, Z., Xiao, Y., Liu, X., Du, Y. & Li,
614 J. 2016. Consumers' behaviors and concerns on fresh vegetable purchase and safety in
615 Beijing urban areas, China. Food Control, 63, 101-109.
616 Dawes, J. 2008. Do data characteristics change according to the number of scale points used?
617 An experiment using 5-point, 7-point and 10-point scales. International journal of
618 market research, 50, 61-104.
619 Dosman, D. M., Adamowicz, W. L. & Hrudey, S. E. 2001. Socioeconomic determinants of
620 health‐and food safety‐related risk perceptions. Risk analysis, 21, 307-318.
621 Douglas, M. & Wildavsky, A. 1983. Risk and culture: An essay on the selection of
622 technological and environmental dangers, University of California Press.
623 Durkheim, E. 1933. The division of labor in society, Macmillan, Simon and Schuster.
624 Figuié, M., Bricas, N., Thanh, V. P. N., Truyen, N. D. & de l'Alimentation, E. S.-E. 2004.
625 Hanoi consumers’ point of view regarding food safety risks: an approach in terms of
626 social representation. Vietnam Social Sciences, 3, 63-72.
627 Figuié, M. & Fournier, T. 2008. Avian influenza in Vietnam: chicken‐hearted consumers?
628 Risk Analysis: An International Journal, 28, 441-451.
629 Green, J., Draper, A. & Dowler, E. 2003. Short cuts to safety: risk and'rules of thumb'in
630 accounts of food choice. Health, risk & society, 5, 33-52.
631 Ha, T. M., Shakur, S. & Pham Do, K. H. 2019. Consumer concern about food safety in
632 Hanoi, Vietnam. Food Control, 98, 238-244.
633 Hale, T. M., Cotten, S. R., Drentea, P. & Goldner, M. 2010. Rural-urban differences in
634 general and health-related internet use. American Behavioral Scientist, 53, 1304-1325.
635 Hall, C. & Moran, D. 2006. Investigating GM risk perceptions: a survey of anti-GM and
636 environmental campaign group members. Journal of Rural Studies, 22, 29-37.
637 Herbert, G., Butler, L., Kennedy, O. & Lobb, A. 2010. Young UK adults and the 5 A DAY
638 campaign: perceived benefits and barriers of eating more fruits and vegetables.
639 International journal of consumer studies, 34, 657-664.
640 Hoi, P. V., Mol, A. P., Oosterveer, P., van den Brink, P. J. & Huong, P. T. 2016. Pesticide
641 use in Vietnamese vegetable production: a 10-year study. International Journal of
642 Agricultural Sustainability, 14, 325-338.
643 Järvelä, K., Mäkelä, J. & Piiroinen, S. 2006. Consumers’ everyday food choice strategies in
644 Finland. International Journal of Consumer Studies, 30, 309-317.
645 Jolliffe, I. T. 2002. Principle Component Analysis, Second Edition, Springer.
646 Kasperson, R. E., Renn, O., Slovic, P., Brown, H. S., Emel, J., Goble, R., Kasperson, J. X. &
647 Ratick, S. 1988. The social amplification of risk: A conceptual framework. Risk
648 analysis, 8, 177-187.
649 Kher, S. V., De Jonge, J., Wentholt, M. T., Deliza, R., de Andrade, J. C., Cnossen, H. J.,
650 Luijckx, N. B. L. & Frewer, L. J. 2013. Consumer perceptions of risks of chemical
651 and microbiological contaminants associated with food chains: a cross‐national
652 study. International Journal of Consumer Studies, 37, 73-83.
653 Knight, A. J. & Warland, R. 2005. Determinants of Food Safety Risks: A Multi‐disciplinary
654 Approach. Rural Sociology, 70, 253-275.
655 Kruskal, W. H. & Wallis, W. A. 1952. Use of ranks in one-criterion variance analysis.
656 Journal of the American statistical Association, 47, 583-621.
657 Lachat, C., Huynh, T. T. T., Verstraeten, R., Nago, E., Roberfroid, D. & Kolsteren, P. 2011.
658 Factors associated with eating out of home in Vietnamese adolescents. Appetite, 57,
659 649-655.

16
660 Larsen, K. & Gilliland, J. 2008. Mapping the evolution of'food deserts' in a Canadian city:
661 Supermarket accessibility in London, Ontario, 1961–2005. International Journal of
662 Health Geographics, 7, 16.
663 Lee, L. E., Niode, O., Simonne, A. H. & Bruhn, C. M. 2012. Consumer perceptions on food
664 safety in Asian and Mexican restaurants. Food Control, 26, 531-538.
665 Liu, P. & Ma, L. 2016. Food scandals, media exposure, and citizens’ safety concerns: A
666 multilevel analysis across Chinese cities. Food Policy, 63, 102-111.
667 McEachern, M. G. & Warnaby, G. 2008. Exploring the relationship between consumer
668 knowledge and purchase behaviour of value‐based labels. International Journal of
669 Consumer Studies, 32, 414-426.
670 Nguyen-Viet, H., Tuyet-Hanh, T. T., Unger, F., Dang-Xuan, S. & Grace, D. 2017. Food
671 safety in Vietnam: where we are at and what we can learn from international
672 experiences. Infectious diseases of poverty, 6, 39.
673 Omari, R., Frempong, G. K. & Arthur, W. 2018. Public perceptions and worry about food
674 safety hazards and risks in Ghana. Food Control, 93, 76-82.
675 Pennings, J. M., Wansink, B. & Meulenberg, M. T. 2002. A note on modeling consumer
676 reactions to a crisis: The case of the mad cow disease. International Journal of
677 Research in Marketing, 19, 91-100.
678 Pulliat, G. 2015. Food securitization and urban agriculture in Hanoi (Vietnam). Articulo-
679 Journal of Urban Research. Special issue 7.
680 Rutsaert, P., Pieniak, Z., Regan, Á., McConnon, Á. & Verbeke, W. 2013. Consumer interest
681 in receiving information through social media about the risks of pesticide residues.
682 Food Control, 34, 386-392.
683 Schroeder, T. C., Tonsor, G. T., Pennings, J. M. & Mintert, J. 2007. Consumer food safety
684 risk perceptions and attitudes: impacts on beef consumption across countries. The BE
685 Journal of Economic Analysis & Policy, 7.
686 Shi, T. 2001. Cultural values and political trust: a comparison of the People's Republic of
687 China and Taiwan. Comparative Politics, 401-419.
688 Siegrist, M. 2000. The influence of trust and perceptions of risks and benefits on the
689 acceptance of gene technology. Risk analysis, 20, 195-204.
690 Slovic, P. 1992. Perception of risk: Reflections on the psychometric paradigm. In Krimsky
691 and D. Golding (ed.), New York, Praeger.
692 Slovic, P. 1993. Perceived risk, trust, and democracy. Risk analysis, 13, 675-682.
693 Slovic, P., Fischhoff, B. & Lichtenstein, S. 1982. Why study risk perception? Risk analysis,
694 2, 83-93.
695 Tavakol, M. & Dennick, R. 2011. Making sense of Cronbach's alpha. International journal of
696 medical education, 2, 53.
697 Van Hoi, P., Mol, A. P. & Oosterveer, P. J. 2009. Market governance for safe food in
698 developing countries: The case of low-pesticide vegetables in Vietnam. Journal of
699 environmental management, 91, 380-388.
700 Verbeke, W. 2005. Agriculture and the food industry in the information age. European
701 Review of Agricultural Economics, 32, 347-368.
702 Vietnam News. 2015. Supermarkets sell dubious veggies. Vietnam News, January 2015.
703 Available:
704 https://vietnamnews.vn/society/265631/supermarkets-sell-dubious-veggies.html
705 [Accessed 05/3/2019]
706 Wang, H., Moustier, P. & Loc, N. T. T. 2014. Economic impact of direct marketing and
707 contracts: The case of safe vegetable chains in northern Vietnam. Food Policy, 47, 13-
708 23.

17
709 Wertheim-Heck, S. C., Spaargaren, G. & Vellema, S. 2014. Food safety in everyday life:
710 Shopping for vegetables in a rural city in Vietnam. Journal of Rural Studies, 35, 37-
711 48.
712 Yang, Z., Paudel, K. P., Wen, X., Sun, S. & Wang, Y. 2020. Food Safety Risk Information-
713 Seeking Intention of WeChat Users in China. International Journal of Environmental
714 Research and Public Health, 17, 2376.
715 Yeung, R. M. & Morris, J. 2006. An empirical study of the impact of consumer perceived
716 risk on purchase likelihood: a modelling approach. International Journal of Consumer
717 Studies, 30, 294-305.
718

18
Highlights:

There are similarities as well as differences in the determinants of risk perception

across the rural and urban region
Heightened risk perception led to the avoidance of particular vegetables
Reducing risk perception will enhance the sustainability of food consumption
Regional characteristics should be considered in designing risk communication
policies
Declaration of interest: None