How do citizens perceive farm animal welfare conditions in Brazil?

The aim of this study is to understand the perceptions of Brazilian citizens about the actual conditions of farm animal welfare in the poultry, beef, and dairy supply chains. To reach this aim, an online survey was conducted. The analysis was based on descriptive statistics and three logistic regressions models. Results of descriptive statistics showed that citizens in Brazil had mostly negative perceptions about the actual conditions of animal welfare in the poultry, beef, and dairy supply chains. Results of the logistic regression models showed that in the poultry and dairy supply chains, citizens with background in agricultural/veterinary sciences, and citizens who reported a higher level of knowledge about these supply chains, were more likely to perceive as bad the actual conditions of farm animal welfare. In the poultry supply chain, citizens who reported previous contact with poultry farms were also more likely to perceive as bad the actual condition of farm animal welfare. In addition, the perception that farmers are mainly focused on the economic aspect of farming and less on animal welfare, the perception that animals do not have a good quality of life while housed on farms, and the perception that animals are not adequately transported and slaughtered, negatively impact on perceptions about the actual conditions of farm animal welfare in the three supply chains. We concluded that a protocol aimed to improve citizens’ perceptions about the actual conditions of farm animal welfare should focus in all phases of the supply chains.


4
89 University of Grande Dourados/Faculty of Management, Accounting and Economics. Before starting 90 data collection, the questionnaire was tested with 20 participants. All the questions were translated to 91 Portuguese.

92
To collect the data, we conducted an anonymous online survey. In a first step, we contacted 93 by phone human resource departments in several universities across Brazil. In this first contact, we 94 explained the purpose of our research, and asked if the department would forward a survey link for the 95 personal e-mail of students, professors and administration staff. Upon acceptance, we sent a follow-up 96 e-mail to human resource departments with the survey link and a brief description of the research, 97 which was then disseminated online for the academic community. Each university disseminated the 98 questionnaire of only one supply chain. We received 1.617 questionnaires of which three were 99 disregarded because they were incomplete. The final number of questionnaires was 728 for the poultry 100 supply chain, 586 for the beef supply chain, and 300 for the dairy supply chain. The data collection 101 took place from November 2016 until December 2017.

Statistical analysis
103 Statistical analysis was conducted in two steps. In a first step, we used factor analysis to 104 reduce the number of items used to represent participants' perceptions about animal welfare. Principal 105 component was used as the extraction method. The criterion to define the number of factors was an 106 eigenvalue greater than one [19]. Items were included in a factor when they presented factor loadings 107 greater than 0.5. Factors scores were generated for subsequent analysis [19].

108
In a second step, we run three logistic regression models. The three dependent variables were 109 participants' perceptions about the actual conditions of FAW on each supply chain. In the original 110 questionnaires, this variable was measured in a Likert scale from 1 to 5 (S1 Table). In order to run the 111 logistic models, we transformed the variable participants' perceptions about the actual conditions of 112 FAW on each supply chain into a binary variable, where participants who answered 1 or 2 were 113 gathered to a bad condition group (Bad:0) and participants who answered 3, 4 or 5 were gathered to 114 regular condition group (Regular:1). We tested the impact of two groups of independent variables: 115 participants' socio-demographic characteristics, and participants' perceptions about animal welfare.
116 The significance level was p<0.05.

118
Descriptive statistics 119 Descriptive statistics of participants' socio-demographic characteristics are presented in S3 120 Table. Socio-demographic characteristics were similar for the participants in the poultry and beef 121 supply chains, but somehow different for participants in the dairy supply chain. Participants who 122 answered the dairy supply chain questionnaire were, on average, older, more educated, and earned a 123 higher income compared to participants who answered the poultry and beef supply chains 124 questionnaires. Apart from these differences, others participants' socio-demographic characteristics 125 were similar within the three supply chains: most participants were female, most of them study/work 126 out of the fields related to agricultural/veterinary sciences, most of them had previous contact with 127 farm animals, most of them lived in urban areas, and most of them were pet owners.

128
Descriptive statistics of participants' perceptions about the actual conditions of FAW on each 129 of the three supply chains and other questions related to animal welfare are presented in S3 Table. In

152
Descriptive statistics about the statements used to measure participants' perceptions about 153 animal welfare are presented in S2 Table. For the statements related to FI (Perc 1 , Perc 2 , Perc 3 , Perc 4 ), 154 the mean were above or close to 4, which indicates that participants agreed that most farmers focus too 155 much on the economic aspect of farming and less in animal welfare. For the statements related to LQ 156 (Perc 5 , Perc 6 , Perc 7 , Perc 8 ), the mean were below or close to 3, which indicates that participants did not 157 agree that animals have a good quality of life while housed on farms. For the statements related to HC 158 (Perc 9 , Perc 10 ), the mean were below or close to 2, which indicates that participants agreed that humans 159 are allowed to use animals for consumption.

160
Logistic regression models 161 We tested the impact of socio-demographic characteristics, and participants' perceptions 162 about animal welfare on their perceptions about the actual condition of FAW in each supply chain.
163 Results of the three logistic regression models are present in Table 1. The socio-demographic 164 characteristics age, gender, pet ownership, and consumption of animal products did not significantly 165 impact on participants' perceptions about the actual condition of FAW in any supply chain. In the 166 poultry supply chain, participants who reported previous contact with poultry farms were more likely 167 to perceive as bad the actual condition of FAW compared to participants who had not reported previous 168 contact. In the poultry and dairy supply chains, participants in the fields of study related to 169 agricultural/veterinary sciences were more likely to perceive as bad the actual conditions of FAW 170 compared to participants out of these fields. In those supply chains, participants who reported a higher

245
A potential limitation of this study concerns selecting participants only in the academic 246 community. In comparison to the Brazilian population our sample is younger, more educated, and 247 earns a higher income [23]. Although we acknowledge that our sample is unbalanced in terms of 248 education, income, and age, we argue that academic community members have more access to 249 information that might drive changes in production systems.