A system-wide approach to managing zoo collections for visitor attendance and *in situ* conservation

The study examines how zoo collection composition and socio-economic context influence two core outcomes: visitor attendance and institutional contributions to in situ conservation. While modern zoos aim to balance conservation, education, research, and entertainment, perceived public preferences for large vertebrates—especially mammals—can conflict with conservation-driven recommendations to focus on smaller-bodied, threatened, or native species and to specialize collections. There is concern that compositional shifts away from large charismatic species could reduce attendance and thus lower financial capacity for in situ projects. Past research indicates attendance is affected by collection popularity and socio-economic factors, but comprehensive global analyses linking composition to attendance and to in situ activity, including direct and indirect effects and trade-offs, are lacking. The authors test these relationships using structural equation modelling across a global sample of zoos, hypothesizing that both collection attributes (e.g., total animals, species richness, body mass, dissimilarity, mammals) and socio-economic variables (local population density, GDP) drive attendance, which in turn affects in situ contributions.

Prior work suggests that zoos attract over 700 million visitors annually and collectively invest substantial funds in in situ conservation, with attendance linked to collection popularity. The flagship species approach posits that large charismatic vertebrates can drive education and fundraising that benefit broader conservation goals. However, large species are costlier to maintain and pose welfare and management challenges, leading to recommendations to prioritize smaller-bodied or threatened species and to rationalize collections. Socio-economic variables (e.g., GDP, population) also influence attendance. Despite these insights, previous studies were geographically or taxonomically limited and primarily assessed direct effects, lacking an integrated framework capturing indirect pathways and trade-offs between composition, attendance, and conservation outcomes. This study builds on that literature by integrating global datasets and SEM to quantify both direct and indirect effects.

Design: Two structural equation modelling (SEM) frameworks were developed: (1) an attendance model for 458 zoos in 58 countries, and (2) an in situ model for a subset of 119 institutions (AZA members across four countries) with reported in situ contributions. An a priori meta-model informed by literature guided model specification. Data sources: Annual attendance and institutional area from the International Zoo Yearbook (IZY, 2015); vertebrate species holdings (mammals, birds, reptiles, amphibians) from Species360 ZIMS (2017); IUCN Red List statuses standardized via the R package taxize; species body mass from the Species Knowledge Index, with missing masses imputed at higher taxonomic levels; socio-economic data: national GDP and population (World Bank, 2017) and estimated local population within a 10 km radius (CIESIN GPWv4, 2017). In situ contributions were the number of field conservation programs supported by institutions (AZA Annual Report on Conservation and Science, 2015). Variables (2017 unless noted): attendance (2015); total animals; species richness; mammal species richness; mean abundance-weighted species body mass per institution; Brillouin diversity (alpha) and Raup–Crick dissimilarity (beta) among collections; threatened species proportion (IUCN CR+EN+VU); institutional area (ha; 2015); GDP (US$; 2015); national population (2015); local 10 km population (2015); in situ contributions (2015). Mean abundance-weighted body mass was calculated as M = (Σ x_i m_i)/(Σ x_i), where x_i is species i abundance and m_i its body mass. Analysis: SEM implemented in R (lavaan, lavaan.survey). Variables were mean-centered and standardized. Model refinement followed Grace et al.-style procedures: evaluate absolute and incremental fit indices, use modification indices (chi-square criterion 3.84) with theoretical justification, remove unsupported paths (p≥0.05), and select models via AICc (>2 units improvement). The attendance model informed overlapping paths in the in situ model due to higher power. Mediation tests assessed direct and indirect effects. Species-abundance weighted models are primary; presence-absence variants were also analyzed and yielded qualitatively similar conclusions with limited differences. Model validation used four random data subsets (n=200 each) for the attendance model. Due to smaller sample size and geography, GDP and country were omitted from the in situ model.

- Attendance and in situ link: Higher attendance strongly predicts more in situ conservation projects (standardized direct effect 0.583; R2 for in situ model = 0.496). - Key drivers of attendance (R2 = 0.689; n = 458): - Total number of animals: strongest positive direct effect on attendance (0.587, p<0.001). - Local population (10 km radius): positive direct effect (0.444, p<0.001). - Mean species body mass: positive direct effect (0.340, p<0.001) but small total effect (0.062) due to negative indirect effects via reductions in species richness, total animals, and dissimilarity. - GDP: positive direct effect (0.163, p<0.001) with a negative indirect effect via fewer total animals; total effect 0.083. - Collection dissimilarity: positive direct effect (0.125, p<0.001). - Mammal species richness: small positive direct effect (0.102, p=0.021), larger total effect (0.309), indicating mammals are especially important for attendance. - Overall species richness: direct negative effect (-0.184, p=0.004) but positive total effect (0.262) mediated by increased total animals. - Determinants of in situ contributions (n=119): - Attendance: strong positive effect (0.583, p<0.001). - Proportion of threatened species: positive effect (0.189, p=0.004). - Institutional area: positive direct effect (0.169, p=0.015) and larger total effect (0.320). - No direct effect of mean body mass on in situ projects; its influence is mediated through attendance. - Internal pathways and trade-offs: - Total animals (R2 = 0.783) increases with species richness (0.759) and area (0.309), but decreases with higher body mass (-0.157 direct; total -0.483) and with GDP (-0.136). - Species richness (R2 = 0.678) increases with mammal species richness (0.790) and area (0.096), and decreases with higher body mass (-0.429). - Dissimilarity (R2 = 0.257) increases with area (0.277) and decreases with higher body mass (-0.593). - Overall pattern: Zoos with many animals, larger-bodied species, higher species richness (especially mammals), and collections dissimilar to others attract more visitors, which in turn is associated with greater in situ conservation activity. A clear trade-off exists between showcasing large-bodied species and maintaining high numbers of animals and collection distinctiveness, implying alternative strategies (e.g., many small, unique species) can also yield high attendance and in situ contributions.

Findings show attendance is the principal conduit linking collection composition to in situ conservation: institutions with more animals, larger-bodied species, greater mammal richness, and more distinctive collections attract more visitors and support more in situ projects. However, mean body mass has no direct effect on in situ contributions, acting indirectly via attendance, indicating that large vertebrates are not intrinsically required for in situ activity if attendance can be maintained through other composition strategies. The positive association between the proportion of threatened species in collections and in situ contributions suggests integration of ex situ focus with in situ actions (aligned with IUCN SSC guidelines and the One Plan approach) can elevate conservation outputs, even though threatened status does not boost attendance. The results support the flagship approach—popular large vertebrates can effectively drive attendance and fundraising—but highlight trade-offs: increasing mean body mass reduces total animals and collection dissimilarity, both beneficial to attendance, presenting an alternative path to success via many smaller, unique species. Ethical, welfare, and management considerations (e.g., enclosure size, population management standardization) may constrain simple attendance-maximizing strategies and can conflict with recommendations to standardize collections. The study underscores the need to account for both direct and indirect effects and system-level trade-offs in collection planning to enhance both visitor engagement and conservation outcomes.

This work provides the first global, system-level evidence linking zoo collection composition and socio-economic context to visitor attendance and, in turn, to in situ conservation contributions. It demonstrates that multiple collection strategies can achieve high attendance and conservation support: (1) exhibiting numerous large-bodied species, particularly mammals, and (2) alternatively, focusing on many smaller-bodied, unique species that increase dissimilarity while maintaining high total animal numbers. Elevating the proportion of threatened species in collections is associated with greater in situ engagement, though not with attendance. The study offers an evidence base for collection planning and policy to balance public expectations, welfare and management constraints, and conservation objectives. Future research should refine measures of conservation impact beyond project counts, expand in situ datasets beyond AZA regions, incorporate financial and welfare metrics, and test causal mechanisms with longitudinal and experimental designs.

- Data provenance: IZY and ZIMS rely on institution-submitted records without centralized editorial verification, potentially introducing inconsistencies in attendance reporting and species holdings updates. - Proxy measures: Attendance used as a proxy for income; number of in situ projects used as a proxy for conservation investment (validated with anonymized BIAZA data but not integrated in SEM). - Sample scope: Attendance model spans 458 institutions in 58 countries; the in situ model is limited to 119 AZA institutions across four countries, and omits GDP and country effects due to sample size. - Taxonomic/body mass data: Body mass missing values imputed at higher taxonomic levels; analysis limited to vertebrates (aquariums and certain centers excluded). - Model assumptions: SEM captures correlations consistent with hypothesized causal pathways but cannot definitively prove causation; results depend on model specification and measured variables. - Trade-offs and welfare: Practical constraints (animal welfare, enclosure size, management policies) may limit the feasibility of composition strategies inferred from statistical associations.