Characterizing bird‐keeping user‐groups on Java reveals distinct behaviours, profiles and potential for change

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2020 The Authors. People and Nature published by John Wiley & Sons Ltd on behalf of British Ecological Society 1Department of Natural Sciences, Manchester Metropolitan University, Manchester, UK 2BirdLife International, Cambridge, UK 3Cornell Lab of Ornithology, Cornell University, Ithaca, NY, USA 4Cedar House, Chester Zoo, Upton-byChester, Chester, UK 5Fakultas Teknobiologi, Kampus II Gedung Thomas Aquinas, Universitas Atma Jaya Yogyakarta, Yogyakarta, Indonesia


| INTRODUC TI ON
Around 5,000 species of terrestrial birds, mammals, amphibians and reptiles are globally threatened with extinction due to overexploitation in the international wildlife trade, and this number may almost double in the near future (Ribeiro et al., 2019;Scheffers, Oliveira, Lamb, & Edwards, 2019). Bird species are far more widely represented in trade than mammals, and a disproportionate number of avian taxa are threatened by overexploitation (Alves, Lima, & Araújo, 2013;Bush, Baker, & Macdonald, 2014). This is particularly prevalent in Southeast Asia (Coleman et al., 2019;Harris et al., 2017), where intense demand has precipitated an 'Asian Songbird Crisis' (Lee, Chng, & Eaton, 2016;Rentschlar et al., 2018;Sykes, 2017).
Halting the extraction of birds from the wild, or at least reducing it to sustainable levels, is thus a global conservation priority (Bezerra, Araújo, & Alves, 2019;Marshall et al., 2020a;Symes, Edwards, Miettinen, Rheindt, & Carrasco, 2018) alongside addressing the problem of habitat loss, which in Asia threatens more bird species than anywhere except Amazonia (BirdLife International, 2020).
The trapping and trading of birds globally is driven principally by demand for pets, but also by the need for nutritional and medicinal resources, symbolic or cultural practices and gambling-related contests (Bezerra et al., 2019;de Oliveira, de Faria Lopes, & Alves, 2018;Jepson, 2010;Harris et al., 2017;Souto et al., 2017). Domestic consumption of birds as pets in two large biodiverse countries, Brazil and Indonesia, may actually be larger than the total international market (Alves et al., 2013;Jepson & Ladle, 2005;Rentschlar et al., 2018).
Regulating domestic trade to prevent significant impacts on wild bird populations is, however, problematic, as the size and variety of the networks involved can make enforcement logistically and politically difficult (Alves et al., 2013;Bezerra et al., 2019).
In Indonesia, where at least 26 bird species are globally threatened through overexploitation (BirdLife International, 2020), most of the trade is domestic (Chng, Eaton, Krishnasamy, Shepherd, & Nijman, 2015;, but demand also drives the importation of birds from other countries in the region (Leupen et al., 2018). The legislation surrounding the trade in wild birds in Indonesia is comprehensive, and the list of protected species, which can only be traded if they are captive-bred, was recently updated to include newly recognized and recently Red-Listed species Miller, Gary, ansyah, Sagita, & Adirahmanta, 2019). Even the harvest of unprotected wildlife is, in theory at least, regulated through a quota system set by a governmental body, the Indonesian Institute of Sciences (LIPI). Harvest quotas have, however, only been set for a few species, thereby rendering the capture or trade of any other species illegal . Nevertheless, the trade and ownership of wild-caught birds is ubiquitous across Indonesia Marshall et al., 2020a) and bird traders are often confused about or unaware of the law (Rentschlar et al., 2018) making enforcement both difficult and unpopular (Janssen & Chng, 2018;Miller et al., 2019).
Initial research explored the underlying behaviours and motivations of bird-keepers from an anthropological or historical perspective, and proposed a market-based way to reduce pressure on wild bird populations (Jepson, 2010;Jepson & Ladle, 2005, 2009Jepson, Ladle, & Sujatnika, 2011). This entailed substituting captive-bred birds under a certification scheme, promoting singing competitions between captive-bred birds only and establishing ringing courses to help distinguish wild-caught from captive-bred individuals (Jepson & Ladle, 2009). Even so, recent evidence indicates that captivebreeding has not been able to meet the demand for songbirds Harris et al., 2015Harris et al., , 2017.
Identifying and characterizing consumers based on behaviours and preferences has allowed researchers to break seemingly homogeneous audiences into groups on which to target demand reduction efforts (Razavi & Gharipour, 2018;Shairp, Veríssimo, Fraser, Challender, & Macmillan, 2016;Williams, Gale, Hinsley, Gao, & St. John, 2018). Such techniques have helped to understand demand for various wildlife products including orchids (Hinsley, Veríssimo, & Roberts, 2015), rhino horn (Dang Vu & Nielsen, 2018;Truong, Dang, & Hall, 2016) and saiga horn , and their potential value for finding ways to reduce demand for Asian songbirds requires urgent exploration.
In this study we seek to distinguish songbird-keeping usergroups on Java based on their behaviours and preferences, and to identify the demographic determinants of user-group membership.
We also track differences in bird taxa owned across user-groups and 5. Efforts to increase the sustainability of bird-keeping in Java should focus on emphasizing the importance of captive-bred birds, in particular to hobbyists, the largest user-group, whose bird-keeping behaviour poses the biggest threat to wild bird populations, whilst also incentivizing legitimate breeding enterprises among contestants and breeders.

K E Y W O R D S
cage-bird, conservation marketing, consumer demand, sustainable use, wildlife trade the degree of movement between user-groups over a 2-year period.
Our profiles of user-groups aim to identify specific threats to wild bird populations by characterizing for each group (a) species typically owned; (b) preferences for wild-caught or captive-bred birds and (c) number of birds owned and turnover of individual birds. This exercise may then benefit conservation by segmenting audiences on behaviour and demographics in such a way as to allow demand reduction interventions to be more appropriately and precisely targeted (Hinsley et al., 2015).

| Study design
In 2018 we collected data on bird ownership characteristics during a survey of households on Java, Indonesia, using a stratified sampling technique to capture a spectrum of rural and urban districts within each of the island's six provinces (Marshall et al., 2020a). Within communities and neighbourhoods of selected districts, households were systematically sampled (full details on sampling methodology can be found in Appendix A), and interviews carried out with the most senior member of the household available.
The motivations for bird-keeping in Java include the desire for success in contests, which drives preferences for birds with high-quality songs or colours , and the desire for social status, which drives preferences for birds that are normally hard to acquire (Jepson, 2016). However, broad user-groups are primarily described in terms of recreational pursuits (Thomas-Walters et al., 2019). The heterogeneity of the bird-owning community  allows us to characterize three potential user-groups: (a) hobbyists, who keep birds primarily as pets and rarely engage in competitions or captive-breeding; (b) contestants, who keep birds primarily to enter them in singing contests, but may also breed birds; and (c) breeders, who breed and/or train birds for resale or as a hobby, but do not regularly enter birds in contests.
To assign bird-keepers to one of the three user-groups, respondents were asked to choose all motivations for keeping birds that were applicable to them: (a) to keep as a hobby, (b) to enter singing contests and (c) to breed or train birds. We also collected data on: species identity, abundance and origin (i.e. captive-bred or wildcaught) of all cage-birds in the household; the consumption behaviour and preferences of bird-keeping respondents (i.e. number and fate of birds owned previously; purchasing habits; time spent tending birds); and socio-economic and demographic profiles at both household and individual levels (see Appendix B for list of survey questions).
To represent household socio-economic status objectively, we used a composite household asset index (HAI: Filmer & Pritchett, 2001). We adopted a checklist of household items and conditions (Schreiner, 2012) and summed the total number of such items to create a score to serve as a proxy for the economic status of the respondent, with higher score indicating greater affluence (Harttgen & Vollmer, 2013). To establish a household occupancy index, we asked respondents how many people lived in their household and how many bedrooms they had, and then calculated the number of people per bedroom. To estimate losses of birds, we calculated the proportion of them owned in 2016 that respondents reported to have subsequently died. As the owning of trafficked wildlife is not illegal under Indonesian legislation  our questions did not relate to perceived illegal behaviour; thus in common with previous research into songbird-keeping (Burivalova et al., 2017;Krishna et al., 2019) we assumed that respondents provided information about the origins of their birds truthfully.
We defined cage-birds as we did in Marshall et al. (2020a)-birds (both native to Indonesia and exotic) kept, bought or sold as pets or used in singing contests, including passerines (Passeriformes), pigeons and doves (Columbiformes), owls (Strigiformes), woodpeckers (Piciformes) and cuckoos (Cuculiformes). When birds owned by respondents were actually seen by interviewers (>80% of survey events), they were, in the majority of cases, identified to species level. When birds were not seen, or the interviewer could not recognize them, identification was based on respondent use of market names for the birds, and almost always resulted in their being assigned only to genus level. For example, several species of leafbird Chloropsis spp. have one common market name, as do white-eyes Zosterops spp. Taxonomy follows del Hoyo and Collar (2014, 2016).

| Analysis
We profiled the three user-groups based on bird-keeping habits, focusing on the differences in prevalence of behaviours and preferences; where appropriate, differences were tested across groups using Kruskal-Wallis and chi-squared tests. We fitted binary logistic mixed effects regression models (GLMMs) to identify those socioeconomic attributes associated with (a) ownership/non-ownership of cage-birds and (b) user-group membership versus non-membership among bird-keepers (explored in three separate models). We excluded responses from households where the principal bird-keepers were not present, except for the initial analysis concerning presence or absence of cage-birds within a household. In all models, community was included as a random factor to account for pseudoreplication across the 92 communities. We used model selection and averaging based on the Akaike information criterion (AIC), creating global models with all potential predictors (Table S1); prior to inclusion continuous variables were standardized and checked for collinearity, and predictors with high variance inflation factors (>1.9) were excluded. The top models were defined as those within ΔAICc < 2 of the model with the lowest AIC value (Grueber, Nakagawa, Laws, & Jamieson, 2011). If no model proved better (i.e. Akaike weight < 0.6) from a top set of candidate models, model-averaging was performed, calculating full (zero) method-averaged parameter estimates and using measures of relative variable importance to determine the strength of a predictor's association with the response variable Random forests, a nonparametric decision-tree-based technique that uses bootstrapped subsets of training data to generate an ensemble of models that are then aggregated into a final model (Breiman, 2001), were used to identify characteristics of user-group membership based on numbers of bird species and individuals and on composition of taxa owned by households in 2018. We used repeated 10-fold cross-validation over a tuning grid of potential values to parameterize the model (i.e. the number of variable splits and trees generated) to achieve the highest predictive accuracy (Kuhn, 2008). The statistical and random forest analyses were carried out using the MuMIn (v1.15.6, Bartoń, 2018), lMe4 (Bates, Machler, Bolker, & Walker, 2015), randoMForest (Liaw & Wiener, 2002) and caret (v6.0-84, Kuhn, 2008)

| Ethics statement
Research teams gained permission from, and agreed to stipulations set by, the heads of neighbourhood and relevant administrative authorities prior to data collection. Interviewers always received prior informed consent from respondents. Name of interviewer and time and date of survey were recorded before interviews; all data were

| Household demographic data
With an interview response rate of ~60% (Marshall et al., 2020a), we surveyed 3,040 households from all six provinces of Java.
Based on Java's reported 2010 census population of 36,720,166 households, the estimates of bird ownership we present have an associated ± 1.68% margin of error at the 95% confidence level (Newing, 2010). A comparison of the demographic attributes of our sample and the 2010 census data is given in Table S2. Median age (lower quartile-upper quartile) of respondents was 42 (16-91). Most respondents had a high school education (60%), and the largest occupational category was manual labour (35%), yet a large minority were not in formal employment (29%; Table S1). The mean ± SD HAI score was 14.8 ± 4.8 (range = 0-34), and the median (lower quartileupper quartile) number of people per bedroom was 1.7, 1-2. Of households surveyed, 957 (31%) kept birds in 2018; of the remaining 2,083 (69%), 1,603 (77%) had never kept birds, while 161 (8%) kept birds in 2016.

| Bird-keeping behaviours
Differences in numbers of birds owned, purchasing habits and time spent tending birds per day were most marked between hobbyists and the two other user-groups (contestants and breeders; Table 1). Hobbyists (57% of bird-keepers) tended to keep only small numbers of individuals and species but high proportions of wildcaught birds. Hobbyists were the most likely to receive birds as gifts, although trapping birds themselves or buying them directly from trappers or travelling salesmen was equally prevalent across all user-groups. Contestants and breeders shared many characteristics, but contestants tended to buy more expensive birds and spend more time tending their birds than breeders. Mortality of birds since 2016 was highest in the hobbyist group (proportion of birds that died was 0.22 for hobbyists vs. 0.13 in contestants and 0.15 in breeders), but the difference was not significant. While all user-groups owned threatened species, hobbyists owned a greater proportion of them than the others. Although there were only small differences in preferences concerning the song quality of wild-caught and captive-bred birds, hobbyists were the least likely to express a preference or to take origin into account when purchasing birds (Table 2).

| User-group classification
Our user-group classification had an overall accuracy of 84% (Table S3) Table S4). Overall, the biggest change between the 2 years was an increase in proportions of hobbyists and contestants, both with relatively large recruitment from non-bird ownership in 2016.

| Socio-economic profiles
Our mixed effect models indicated the importance of seven demographic and geographic variables in characterizing cage-bird ownership, and subsequently user-group membership (Figure 3; full model outputs in Table S5). Compared to those who owned no birds ('non-bird-keepers'), bird-keepers were more likely to live in urban communities and in the eastern provinces. They were also more likely to be employed, and to have attained a high school education, while non-bird-keepers were more likely to have experienced TA B L E 1 Characteristics and preferences of the three songbird-keeping user-groups (respondents self-reported membership of these groups). n varies according to numbers of disregarded responses for various questions, the lower number of people keeping birds in 2016 and reluctance to answer. n was particularly low for losses of birds: hobbyists n = 213, contestants n = 154 and breeders n = 103. Differences in numbers of birds owned and money and time spent on birds were tested using between-group post hoc differences from Kruskal-Wallis, the remainder with χ 2 tests (e.g. H < C indicates hobbyists showed a significantly lower response than contestants)    Table S6), occupation (contestants were the most likely to be employed in business); and demography (hobbyists tended to be older than both breeders and contestants, who were the youngest user-group; Figure 3).

| D ISCUSS I ON
The clearest and most significant threat to wild bird populations from bird-keeping is the consumption behaviour of Java's most abundant user-group, hobbyists, which may represent up to seven million households (Marshall et al., 2020a). The high volume of birds owned by this group, including the largest proportion of potentially wild-caught and threatened birds, is acquired primarily through convenience and availability, with little importance placed on origin or song quality (Burivalova et al., 2017). Furthermore, mortality of cage-birds was highest among hobbyists, and the sheer numbers of hobbyists keeping wild-caught birds across Java means that there is likely to be a huge throughflow of birds into the market . Conversely, the prevalence (Marshall et al., 2020a) and abundance of highly sought-after taxa (e.g. White-rumped Shama, Oriental Magpie-robin, leafbirds) kept by contestants suggests that an anthropogenic Allee effect (Courchamp et al., 2006) is at work, drawing some species into an extinction vortex through their ever-increasing rarity in the wild, market value and status-giving properties Krishna et al., 2019). Although breeders show similar behaviours and preferences to contestants, they also favour profitable taxa (lovebirds, canaries Serinus spp., doves) that can be easily bred and resold for a much-elevated price. Indeed, the capacity for F I G U R E 3 Effect sizes (with 95% CIs) of the (a) geographic, (b) occupational and (c) demographic predictor variables with the highest relative variable importance (>0.6) across models predicting bird ownership (against non-bird ownership) and user-group membership (against other bird-keepers) contestants and especially breeders to produce their own birds may offer a counter to trapping pressures on wild populations (Nijman, Langgeng, Birot, Imron, & Nekaris, 2018). Nevertheless, an unknown but potentially significant proportion of birds held by bird-keepers in Java may come from low-intensity recreational trapping in the wild. Moreover, the large numbers of birds kept, predictably high mortality of wild-caught birds during capture, transportation and marketing (Indraswari et al., 2020) and low survival of many sensitive species in captivity, combine to suggest that the drain on wild populations is likely to be high.

| Informing evidence-based behaviour change
Our study sought to profile songbird-keeping user-groups by characterizing and identifying the behaviours that should underpin conservation efforts to increase the sustainability of birdkeeping. In combination with previous studies, we are closer to understanding the temporal dynamics of demand for songbirds and the implications these pose for future conservation efforts (Jepson & Ladle, 2009;Marshall et al., 2020a). Bird-keeping has increased in prevalence in urban centres in Java, and the abundance of captive-bred exotic birds, such as lovebirds and canaries, has grown dramatically (Marshall et al., 2020a). Tracking changes in behaviours, and in particular those that have the largest impact on wildlife populations, is vital to determining the success of conservation interventions (Veríssimo & Wan, 2018).
This study contributes to the body of evidence on Indonesian songbird-keeping practices by expanding the detail of how usergroups differentially effect bird populations, establishing a baseline against which interventions aimed at reducing the impact on wild birds can be measured (Reddy et al., 2017). Previous efforts to increase the availability and popularity of captive-bred alternatives (Jepson & Ladle, 2009) have unfortunately been neutralized by a large increase in the prevalence of often wild-caught native birds (Marshall et al., 2020a). Future efforts should focus on the 'demarketing' (Veríssimo, Vieira, Monteiro, Hancock, & Nuno, 2020) of wild-caught birds in addition to redirecting demand (Moorhouse, Coals, D'Cruze, & Macdonald, 2020) towards captive-bred birds among all user-groups, but hobbyists in particular. Given that effective behaviour change usually requires considerable time (Greenfield & Veríssimo, 2019), movement between user-groups even over a very short (2-year) period could reduce the chances of targeted interventions having a lasting effect on their behaviours and preferences. On the other hand, this dynamism may reflect a responsiveness and flexibility among the population towards adopting more sustainable birdkeeping behaviours. Demand reduction campaigns certainly need to operate on this latter assumption.
A key intervention to reduce demand for wildlife products is the dissemination of information and targeting of campaigns (Veríssimo, Challender, & Nijman, 2012). The bird-keeping community in Java could represent as many as 12 million households (Marshall et al., 2020a). By breaking down this vast audience into user-groups the possibility arises of tailoring and targeting messages for their maximum impact. Interestingly, bird-keepers tended to have moderate levels of education, with our result suggesting that there may be at least two separate non-bird-keeping groups based on educational attainment, those who have not achieved a high school education and those who have achieved higher levels of education. Slightly more affluent, hobbyist bird-keepers are typically middle-aged and from the western provinces, so increasing the importance placed on the origin of birds, as well as on the quality and longevity of captive-bred individuals (Burivalova et al., 2017), may help stem the large inflow of wild-caught birds into hobbyist households. Aspects of bird-keeping have moved away from traditional practices (Jepson & Ladle, 2009) as evidenced by the younger, urban profile of contestants which, as a key consumer demographic in driving national business, suggests competitive bird-keeping will remain an important aspect of the Indonesian economy (Naafs, 2018). Consequently, the choice and source of taxa for competitive bird-keeping among Java's young urban men must be key targets in any campaign to achieve sustainability in the bird trade. Breeders, however, appeared to be the least likely to stop bird-keeping in the short term, more often becoming contestants and less often hobbyists. It may be that, as the most invested group, breeders frequently change the species they keep, both influencing and reacting to market trends; if so, they may be receptive to conservation programmes promoting the captive-breeding of threatened species.
The greater financial and temporal investments made by contestants and breeders in their birds, which acquire both status-earning and resale value, may help explain why bird origin was more important for them than for hobbyists. There is huge potential profit and status in breeding and training birds , and initiatives could stress the value to be placed on origin (equivalent to 'pedigree'). Contestants and breeders both stressed the importance of sourcing birds from particular locations, and promoting a strong cultural attachment to place (Kristianto & Jepson, 2011)  Nevertheless, successful conservation marketing campaigns and environmental education can shift social norms and increase compliance with local legislation (Salazar, Mills, & Veríssimo, 2019;Veríssimo & Wan, 2018). In view of the importance placed on community responsibility and legislation (Kristianto & Jepson, 2011) conservationists could borrow from such approaches to highlight the social undesirability, illegality and risks associated with the laundering or trapping of birds.

| Limitations and caveats
We sought to obtain as representative a sample as possible of households across urban and rural districts from all six provinces of Java by combining a stratified sampling approach to district selection (Marshall et al., 2020a) with the systematic sampling of households within selected districts. When comparing the demographic profile of our study sample with available data from the 2010 Indonesian Census (Badan Pusat Statistik, 2010) for Java as a whole, there are some differences in a number of attributes (see Table S2 in Appendix B). Overall, our sample under-represented those aged 15-24 (14% less than the census), those who have achieved a degree or higher educational attainment (17% less) and those who live in smaller households (21% less), and over-represented those who have achieved high school education (15% more; Table S2). These differences suggest our approach had some of the limitations of previous research (Jepson & Ladle, 2009). For example, there are difficulties in obtaining access and research permissions from certain gated communities that typically occur in more affluent urban areas. The potential bias the omission of such communities creates may be accentuated by their importance in driving trends in the consumption of rarer highly prized species among portions of the bird-keeping community (Jepson, 2016). Future work should address this issue, potentially using online survey techniques to reach such 'high end' consumers (Baltar & Brunet, 2012;Bornstein, Jager, & Putnick, 2013).

| Conclusions
Although conservationists may justly view bird-keeping as inherently detrimental to wild bird populations (Sykes, 2017), within Indonesia the trade in birds is seen as far too economically important and culturally ingrained to be halted completely (Jepson, 2016). Moreover, despite the accumulating evidence of rolling local and even global extinctions , the long tradition of breeding native species (such as Zebra Dove) means that commercial breeding is repeatedly identified as a viable solution to the extraction of wild birds (Nijman et al., 2018). Further research is required to define audiences more precisely, explore the attitudes and perceptions of birdkeepers and frame content aimed at changing specific behaviours (Kidd et al., 2019), but our current breakdown into three user-groups offers an opportunity to begin programmes targeting each group.

ACK N OWLED G EM ENTS
We greatly appreciate the generosity of all the respondents who agreed to be interviewed for this research, including those who participated in the pilot study, and we thank V. de Liedekerke and D. Atma Jaya Yogyakarta as the named partner institution. We thank all the students and graduates who assisted with data collection, and also all the local government employees who granted permission to carry out research across the diverse communities of Java. Thanks are also owed to Paul Jepson and two anonymous reviewers for their helpful suggestions and comments that greatly improved the paper.

This research was funded by Chester Zoo, Manchester Metropolitan
University and Oriental Bird Club (OBC). Icons used for Figure 4 were originally made by Freepik, Good Ware and DinosoftLabs from www.flati con.com.

CO N FLI C T O F I NTE R E S T
Nothing to declare.

F I G U R E 4
Profiles for each user-group based on key behaviours and preferences, demography and dynamism, and the potential issues and solutions to reduce the pressure their behaviours place on wild bird populations manuscript. All authors contributed critically to the drafts and gave final approval for publication.

DATA AVA I L A B I L I T Y S TAT E M E N T
Due to the personal nature of the demographic information collected for this study, fully anonymized data are available from the