An overview of reviews (also called meta-review, review of reviews, umbrella review or meta-meta-analysis) is a synthesis over multiple systematic reviews (Hunt et al. 2018). Commonly used in medical and health research, it is still rarely exploited in ecological and environmental research. Unlike in medical research, meta-analyses in the topics we address typically lack a standardized null model (placebo treatment) and have much more variation in the exact questions addressed as well as covariates employed, hampering direct comparisons across meta-analyses. However, overviews of reviews could be useful in biodiversity conservation, for instance, by providing answers to the long-debated question about the usefulness of surrogate species as conservation tools, and we consider the method worth examining.

Our study had two perspectives, namely WTP and biodiversity surrogacy power, and we conducted separate literature searches for each. First, we searched for papers that contained quantitative meta-analyses on individuals’ willingness to pay for conservation, addressing either species, ecosystems, or biodiversity. We used the search strings given in Suppl. material 1: Online Appendix, page 3 in EBSCO, ISI Web of Knowledge and Scopus. Searches were not restricted by disciplines due to the multidisciplinarity of the topic, or by time. We looked for analyses that could contribute to our understanding on conservation donation behavior. We initially screened the papers for relevance based on their titles and abstracts, and then evaluated the remaining ones more closely, excluding e.g., papers that sought to estimate total monetary value of ecosystem services or that were focusing exclusively on use values. Some studies included ecosystem services but distinguished between them and potential flagship types or between use and nonuse values, allowing us to transfer values to our flagship examples.

In total, we found 12 meta-analyses related to willingness to pay that fulfilled our criteria. (Table 1). The typical number of meta-analyses (or reviews) included in an overview of reviews varies across disciplines and is dependent on the number of meta-analyses conducted on the subject. The number of meta-analyses we found was quite similar to, for instance, an overview on forest conservation policies (Börner et al. 2020) or an overview of reviews examining the WTP for public health measures (Costa et al. 2019).

Then, correspondingly, we searched for reviews and meta-analyses of previous surrogate studies evaluating the biodiversity surrogacy power of different taxa used as umbrella, keystone and indicator species from EBSCO, ISI Web of Knowledge and Scopus databases as well as Google Scholar (see a list of search strings in Suppl. material 1: Online Appendix, page 3). We included all taxa regardless of their potential flagship status, as there is no way of knowing whether they have been used as such somewhere, and because our purpose is also to evaluate use of less typical potential flagships. We concentrated on surrogates that can be associated with quantifying biodiversity either as 1) surrogates for representing other species in conservation planning (biodiversity indicators and umbrellas) or 2) species which are important for the existence of other species (keystones). We excluded instead indicator species for the state of the environment (e.g., those used to study the pollution load of aquatic ecosystems). Altogether, 34 papers evaluating biodiversity surrogacy power met our criteria and were included into the overview of reviews (Table 2).

Due to a relatively small number of highly heterogeneous reviews, we did not attempt to apply statistical meta-analyses to the samples of meta-analyses. Instead, we summarized the findings of previous meta-analyses, systematic reviews and essays in Tables 1 and 2. In the case of WTP-meta-analyses, we reported the following variables obtained from the original studies to Table 1: extent (i.e. geographical scope), number of observations (i.e. number of original studies included in meta-analysis), r² (goodness of fit of the meta-regression model), benefit transfer error estimates, significant covariates in the meta-regression model, flagship type (species, ecosystem or other holistic concept) and the identity of the flagship. In addition to this, we compared results from the different meta-regression WTP models by calculating benefit transfer predictions for example cases of flagship types (Table 1).

Meta-analytic benefit transfer means using the meta-regression model to predict a welfare estimate in a new context (interpolation or even extrapolation) and is increasingly used in environmental economics to inform policy making in circumstances that have not been directly studied (Bergstrom and Taylor 2006; Nelson and Kennedy 2009; Rolfe et al. 2015). In our case, it means predicting how much a person would be willing to pay for a flagship, e.g., a threatened charismatic mammal, based on each suitable model from the identified studies, and comparing these predictions. The success of benefit transfer depends on the predictive power of the meta-regression model, and on the significance of the variables of specific interest. Often meta-analyses that aim at BT applications give an estimate of BT error, which is simply the mean or median of differences between predicted and original value for each observation in the raw data (Kaul et al. 2013). Transfer errors of 30–40% have been proposed as acceptable (Navrud and Ready 2007), and a median rate of 39% was reported by Kaul et al. (2013) across 1071 errors from published studies.

To make a BT prediction, one must decide what variable values to input to the model. To make our predictions for WTP for flagship types as comparable as possible with each other, we set as many variables as possible to be equal across the studies, and used sample means for variables unique to specific studies, when available. For example, for all studies where elicitation technique was significant, we used Dichotomous choice, and for three studies that had Gross domestic product (GDP) as a covariate, we used the same value (see Table 1 and Suppl. material 1: Online appendix for all covariate values used). We made all transfer estimates for the same year (2006), which fell within the ranges of most MAs and thus kept our transfer estimates mainly as interpolations (see the footnote in Table 1 for exceptions). It was not possible to standardize geographical location or extent, for example, because there were meta-regressions entirely based on data from even just one country or one continent (Table 1). Another aspect that merits detailed examination is the scope sensitivity, which we considered for the WTP-meta-analyses. The scope sensitivity refers to the degree in which WTP results are sensitive to the amount of good being valued (i.e., how “much” is conserved with the money).

In relation to studies on biodiversity surrogacy power, each paper was screened for the variables listed in Table 3 and their results were collected from papers to Table 2, where applicable (e.g., statistical tests were applicable only for meta-analyses). Some of the reviewed studies examined cross-taxon congruence, and thus investigated whether one taxon can explain the species richness of other taxa. Cross-taxon congruence had three levels: species richness, composition, and complementarity. For these studies, we indicated in Table 2 which levels each study covered. Due to the heterogeneity of the reviewed studies, we were not able to make statistical analyses of theses collected data, but we examined the results qualitatively.

We summarize the results from the 12 meta-analyses that fulfilled our criteria for overview in Table 1. Even though donation behavior in the context of ENGOs’ fundraising has not been the main focus of any meta-analysis and relatively few of the primary studies they analyze, we assume they still provide reasonable guidance on relative differences in preferences between different kinds of species and other flagship types and on the factors contributing to donation propensity.

The samples have been relatively small and not at all representative of different kinds of species. WTP studies on species have been strongly biased toward mammals, somewhat toward birds, while showing only marginal interest toward other taxa, e.g., reptiles and invertebrates (Table 1). In addition, a few meta-analyses had examined willingness to pay for conservation of ecosystems. A geographical bias was also often present, with the emphasis being on developed countries (Jacobsen and Hanley 2009; Hjerpe et al. 2015), or more specifically either on Europe (Lindhjem 2006; Nijkamp et al. 2008; Žáková Kroupova et al. 2016) or North America (Martín-López et al. 2008; Richardson and Loomis 2009; Ojea and Loureiro 2011).

There was great variation in the explanatory power reported for the meta-analysis models, R² values ranging from 8% (Jacobsen and Hanley 2009) up to 90% (Barrio and Loureiro 2010; Subroy et al. 2019), and the number of significant covariates in the meta-regression models varied greatly (Table 1). Model performance can be further evaluated via benefit transfer (BT) errors, which refers to the difference between predicted and original values for data points in the meta-regression models. The BT errors reported in the original meta-analyses are given in Table 1. Benefit transfer was not the purpose of most studies we reviewed, and only three studies assessed transfer errors directly (Richardson and Loomis 2009; Lindhjem and Tuan 2012; Subroy et al. 2019). One study provided an example of how to conduct BT but did not estimate errors (Amuakwa-Mensah et al. 2018), while some just discussed the potential for the use of BT (Lindhjem 2006; Barrio and Loureiro 2010). The reported transfer errors were mostly within or very close to acceptable limits (Table 1), with the exception of wetlands, where the authors of the paper assumed that the used valuation methods had possibly caused a bias in the meta-analysis (Lindhjem and Tuan 2012).

Typical to WTP meta-analyses, there were from one to several significant methodological variables in each model (Table 1 and Suppl. material 1: Online Appendix). Such variables included e.g., valuation method, payment vehicle, elicitation method and timing of payment, but because methodological variables were not consistent across meta-analyses, they could not be fully standardized for our BT predictions (Suppl. material 1: Online Appendix, section 3.2). In addition, relatively few socioeconomic and geographic variables were available, among which country-level GDP per capita had a positive effect on WTP in four MAs (Jacobsen and Hanley 2009; Barrio and Loureiro 2010; Ojea and Loureiro 2011; Hjerpe et al. 2015), household income in one MA (Lindhjem and Tuan 2012) and developing country status had a negative effect in another MA (Amuakwa-Mensah et al. 2018). It would be unwarranted to extrapolate to a typical western household income level from a model based on developing country data only. Nonetheless, for the sake of example, we calculated WTP by substituting the sample mean household income of USD 14 318 in Lindhjem and Tuan (2012) with the US mean household income for that period, USD 55 000 as an additional example. We obtained WTP sums of €91.55, 7.43 and 20.47 to mammals, turtles, and other species, respectively, being much closer to results from other MAs focused on western countries (Table 1). Geographic regions were also considered in some cases, but with inconsistent variables (Suppl. material 1: Online Appendix).

We calculated benefit transfer estimates for the WTP values for the different flagship example cases, which are given in Table 1 (column: “Predicted WTP in € (2019)”). Majority of the values are in the magnitude of some tens of euros, but the values extend from near zero to near €2000. The high outlier value for mountain ecosystems in Žáková Kroupova et al. 2016 could be due to estimating WTP at the daily level, amounting to a very large annual total. The function of Martín-López et al. (2008) produces WTP estimates that are an order of magnitude smaller than the means from their raw data. This limits its use for benefit transfer purposes in absolute terms, but it is nonetheless useful for relative comparisons among flagship types.

Because we were not able to fully standardize the variables for BT, we focus mostly on within-study comparisons of flagship types. Few consistent patterns emerge from the predicted values. Aquatic ecosystems appear to have higher WTP than forests in all meta-analyses where they were compared against each other (Nijkamp et al. 2008; Ojea and Loureiro 2011; Hjerpe et al. 2015). Wetlands received even higher values, but the result in Nijkamp et al.’s (2008) study is based on only six studies and the result of Lindhjem and Tuan (2012) was not considered robust by the authors due to a bias in the valuation methods used. Marine mammals tend to receive higher WTP values than terrestrial mammals (Martín-López et al. 2008; Richardson and Loomis 2009; Subroy et al. 2019). Most taxa received variable WTP predictions, and the order of attractiveness within one study could be opposite to that of another study. For example, reptiles received both larger (Amuakwa-Mensah et al. 2018) and smaller (Martín-López et al. 2008) values than fish, and birds had both larger (Richardson and Loomis 2009; Amuakwa-Mensah et al. 2018) and smaller (Martín-López et al. 2008) values than mammals.

The results regarding species versus holistic concepts as flagships were few and mixed: In one study mammals attracted 3–5 times higher payments than marine or terrestrial habitats (Lindhjem and Tuan 2012), while the broader category of species received lower values than both forests and water or other ecosystems in the study by Ojea and Loureiro (2011). Nijkamp et al. (2008) found highest WTP values for endangered species and landscapes, lowest for national parks, reserves, and wildlife, thus having both holistic and species concepts at each extreme.

The affective aspect that was assessed most often in the meta-analyses was species’ charisma. Its impact on WTP appeared positive in three of the MAs (Martín-López et al. 2008; Richardson and Loomis 2009; Subroy et al. 2019). However, the results of Amuakwa-Mensah et al. (2018) imply a more complex relationship of charisma with other factors: they found an interaction between charisma and endangerment such that for threatened species charisma increased WTP but for endangered species charisma reduced WTP. Charisma is indeed often contrasted with scientific criteria. In their meta-regression Martín-López et al. (2008) found no effect for any scientific criteria (IUCN threat status, endemism, and ecological role), and only anthropomorphism (eye size) as an indicator of charisma and taxon (mammal > bird > fish > reptile) increased WTP, while negative economic impact (e.g., species associated with human-wildlife conflicts) of the species decreased WTP. Subroy et al. (2019), instead, found that while both had a positive effect on WTP, the impact of threat status was over twice as strong as that of charisma. Barrio and Loureiro (2010) did not examine the impact of threat but found that WTP for forest conservation was smaller in countries with abundant forest cover, possibly reflecting a preference for locally rare aspects. Another potential affective factor influencing WTP for conservation was the locality of the good (the good being the target of payment in the WTP study), which Lindhjem (2006) explored and found a positive association between regional goods and WTP, as opposed to local and national goods.

Scope sensitivity, which refers to the amount of the good being valued (in this case, “how much” is being conserved in terms of e.g., area or population size) may affect the WTP results. Quantifying achievements in conservation is, however, not straightforward. In some of the meta-analyses we reviewed, the area was found to be non-significant (Lindhjem 2006; Barrio and Loureiro 2010). In another study scope-sensitivity was found for absolute changes in area, population size or other measurable unit, but it was absent when change was expressed in relative terms (Ojea and Loureiro 2011), implying it depends on how scope is measured. Hjerpe et al. (2015) found scope sensitivity to be a critical determinant of WTP for forest and wetland ecosystem conservation: Programs with stronger influence on multiple attributes had a predicted WTP up to almost six times higher than for programs where an impact was evident in only an individual attribute. Also, Richardson and Loomis (2009) found a positive relationship of WTP with change in population size.

In some studies, WTP appeared to be larger for avoiding losses of existing values than it was for achieving gains (Lindhjem 2006; Ojea and Loureiro 2011; Subroy et al. 2019) or for restoring degraded environment (Hjerpe et al. 2015), while one study found mixed (Ojea and Loureiro 2011) and another study found the opposite pattern, with higher WTP for securing gains as opposed to avoiding losses (Barrio and Loureiro 2010). In the analysis of Subroy et al. (2019), even though loss avoidance attracted on average higher funds than achieving gains, the magnitude of gains was still positively associated with WTP.

The different types of flagships themselves also represent increasing scope in one dimension, from a single species to multiple, and further to ecosystems or to biodiversity in its entirety. We found no evidence of coherent preference of holistic concepts over single species across meta-analyses, or vice versa. The issue of single vs. many species has been approached in diverse primary studies in different ways, with highly mixed results, but it has not been the specific subject of any meta-analysis, even though Hjerpe et al. (2015) found the number of attributes, which included species, had a strong positive influence on WTP.

Altogether, we identified 34 papers including fourteen meta-analyses, five systematic reviews and fifteen essays for the overview of reviews on biodiversity surrogate types (See Suppl. material 1: Table A1 in Online Appendix). From these 34 studies seven found support for the biodiversity surrogacy power, 23 resulted in mixed or limited findings, three studies found no support and one was unable to draw conclusions (Table 2). Nine meta-analyses and one essay examined the cross-taxonomic congruence, most often from the species richness perspective, but also composition and/or complementarity. They mainly found only partial or limited support for using one taxon as a surrogate (Table 2). These cross-taxon congruence analyses usually explained 14–20% of the variation, Mellin et al. (2011) making a positive exception to this (30–54%, Table 2). The methods used to evaluate cross-taxon congruence varied between meta-analyses as did the interpretation of the results. For instance, Castagneyrol and Jactel (2012) concluded that plant diversity could serve as a surrogate for animal diversity based on correlations between plant and animal species richness (mean r=0.45). On the other hand, for the interpretation of correlations, r>0.70 has been presented as evidence for surrogacy power (Heino, 2010). In addition, two meta-analyses did not find support for the cross-taxon congruence due to the variability in the distribution patterns (Westgate et al. 2014) and variability in concordance between biological groups (de Morais et al. 2018). Although a later study by Westgate et al. (2017) found support for the cross-taxonomic congruence, with plants and birds having the most potential to act as surrogates for other taxa, none of the taxa performed well as surrogates simultaneously for both richness and composition. In another study, birds performed better as surrogates for mammals than for other taxa (Eglington et al. 2012).

According to our overview of reviews, trees can be promising keystone species by providing a variety of resources to other species (Prevedello et al. 2018; van der Hoek et al. 2019) and cushion plants may act as keystone species, but also promote the spread of exotic species (Arredondo-Núñez et al. 2009). Some frugivorous mammals can have potential as keystones through seed dispersal (Stoner et al. 2007). Similarly, beavers can act as keystone species (or ecological engineer) through their various habitat-generating activities (Table 2), but their impact may vary, for example, by region (Rosell et al. 2005). Support for the keystone role of prairie dogs and salamanders was found in one qualitative meta-analysis for each species (Table 2). Instead, the potential umbrella role of large carnivores is less clear and has not been assessed using meta-analytic techniques. Essays assessing this issue have found only partial support for the umbrella role of large carnivores (Noss et al. 1996; Linnell et al. 2000; Sergio et al. 2008). Additionally, large carnivores together as a guild could potentially have a keystone role, but their use as an umbrella or indicator species did not receive full support (Dalerum et al. 2008).

Based on our overview, no conclusions could be drawn for universal effectiveness of either species or holistic flagships in fundraising. A possible reason for this is varying and context dependent preferences of potential donors. Also, past research on donor preferences suggests that potential donors should indeed not be treated as a single group due to their varying values and motivations to donate (Jacobsen et al. 2012; Lundberg et al. 2020). One way to account for this heterogeneity is to use a segmentation strategy to divide donors into smaller groups, which further underscores the need to understand donor preferences for different types of conservation targets.

While some potential donors still prefer the classical charismatic figurehead species, for others different determinants of WTP may be more important (Lundberg et al. 2020). It seems possible that charisma has more impact when respondents have no knowledge of other criteria that could otherwise make a difference. For example, even though charismatic species attracted most funds, respondents reported ‘conservation need’ as the main criterion for choosing species in a WTP survey of Finnish online donors (Lundberg et al. 2019). The lack of effect of the threat status observed in the meta-analysis of Martin-López et al. (2008) could be because the variable was not derived from the original publications but assigned in the meta-analysis, and the respondents may have had varying awareness of the status. In addition, truly non-charismatic species were underrepresented in the evaluated literature, for which the conservation status could have made a larger difference.

Apart from aesthetic or other affective characteristics of species, a further motivation for preferring species over holistic concepts could be so called ‘impact philanthropy’ (Duncan 2004; Amos et al. 2015): certain people prefer to donate to targets where a perceivable impact is more likely; for example, they would rather sponsor a single child through school rather than donate books to an entire school. Wildlife adoption programs, preference for species over holistic targets, or even preference for specific habitats rather than e.g., wilderness as in Jacobsen and Hanley (2009), could all be related to the concept of impact philanthropy. ‘Compassion fade’ is a related phenomenon that can explain these results among people less engaged with environmental issues: It is easier to empathize with an individual species than with broader entities (Markowitz et al. 2013).

Scope sensitivity has been suggested to affect WTP but has been criticized as a poor measure for complex environmental goods (Lindhjem 2006). Defining scope in terms of quantitative (linear) population or protected area gains could also be seen as somewhat problematic in the context of conservation, because even a small improvement for a critically endangered species has a disproportionate importance as opposed to perhaps larger gains for less threatened species in terms of avoided extinctions. Some measure of impact or effectiveness instead of mere area would appear to increase WTP, being in line with the recently described social movement of ‘effective altruism´ (Freeling and Connell 2020), where ‘evidence-based giving’ is assimilated with investing for greatest benefit. While other factors driving preferences (e.g., aesthetics, familiarity, use value, existence value) may obscure potential existing scope effects (Heberlein et al. 2005), it should be noted that economic theory may not be the best predictor for behavior in such value-dependent context to begin with. For example, Jacobsen et al. (2012) identified in a choice experiment three groups of scope-insensitive respondents: (1) those driven by existence value, who prefer to save species from extinction, but see no additional value in larger population sizes (2) those indifferent to the magnitude of increase (genuinely scope-insensitive) and (3) those who may avoid too large increases possibly due to other moral or cost concerns.

It should be noted that it was not possible to fully standardize the BT predictions due to the inconsistent covariates and heterogeneity of original studies. Especially the varying geographical extents of the studies caused problems with the benefit transfer. Even though the WTP values from different countries are purchasing power parity (PPP) corrected in the meta-analyses, benefit transfer to an entirely common setup across meta-analyses is not possible as the studies may or may not have e.g., region, developing country status or household income as covariates. PPP does not explain behavioral variation due to economic and cultural differences (see Ready and Navrud (2006) for challenges of international BT, including discussion on cultural differences, and Kaul et al. (2013) for the importance of geographic variables in BT).

The large variability both in the R² values (i.e., the goodness of fit of the meta-regression model) and the varying number of significant covariates in the meta-regression models (cf. Table 1) suggests that in some cases the variables may have been too coarsely classified, important variables may have been missing, or simply data points have been too few to obtain significant results and high explanatory power. In several cases the results may have been driven by sampling or study design issues. For example, some past results on higher WTP for charismatic species may have been inflated by WTP from ecotourism studies valuing the viewing of species (e.g., Lindhjem and Tuan 2012). Even though such funds could be used for conservation, the motivation behind these choices is the desire to personally experience the species or the location. Such choices are not necessarily related with general conservation donation behavior. WTP for viewing species has been found to vary by over two times even within the Big five species (which include rhinoceros, leopard, African elephant, and the Cape buffalo (Van Tonder et al. 2013)). Most of them are generally considered to be among the most charismatic species (Albert et al. 2018). The higher WTP for regional goods, as opposed to local and national goods observed in Lindhjem’s (2006) study, is not directly in line with the distance decay hypothesis in philanthropy, where people are assumed to prefer local causes (Hanley et al. 2003) but does not contradict it either as the lack of significance for local and national goods could simply be due to data heterogeneity, and lack of international scale. A further example of possible impact of study setup on WTP are Barrio and Loureiro 2010 findings on the lower willingness to protect forests in more forested countries. This could be due to, for example, forests being perceived as something common and less in need of conservation than more rare habitat types (see also Jacobsen et al.’s 2012 first group of scope-insensitive respondents in the paragraph above driven by existence value), or the larger economic importance of commercial forestry in these countries, which can lead to a conflicting interest with conservation (Niemelä et al. 2005).

Another equivocal factor that influenced the BT predictions was the mixed impact of study year on WTP in some of the meta-regression models, in most cases with more recent studies producing lower estimates, which has been attributed in the past to methodological developments in WTP studies (Ojea and Loureiro 2011; Lindhjem and Tuan 2012; Hjerpe et al. 2015), but with the opposite trend in Richardson and Loomis (2009), being in line with increasing environmental awareness. There was no evident explanation for these different conclusions. All of this emphasizes the importance of standardizing primary research efforts, as well as how meta-analyses are conducted, and that interpretations of WTP in absolute terms must be cautious.

The idea of using simplistic taxon surrogates for conservation planning is attractive because data available for making decisions is scarce. However, typically, the theoretical basis behind the expectation of finding surrogate relationships has been poorly, if at all, identified (Sætersdal and Gjerde 2011). Our overview of reviews further highlights this tenuous basis. Based on their meta-analysis Westgate et al. (2014) propose that because of inconsistent results, researchers should clearly articulate the reasons for testing the chosen taxa. They also suggest that future efforts should focus on combinations of taxa (also proposed by some other studies in our meta-review, Table 2), rarely studied taxa, as well as traits of taxa to help identify effective surrogates. The latter idea was exemplified, for instance, by Meurant et al. (2018), who chose species as surrogates for connectivity conservation based on traits that represent vulnerability to fragmentation.

The limited support for cross-taxon congruence in species richness, composition and/or complementarity in our overview of reviews on surrogacy meta-analyses suggest that the usefulness of one taxon as a substitute for others may be limited (Table 2). However, there are methods that still could be tested to identify effective surrogates. The use of large datasets and GIS-based methods has been proposed as a requirement for identifying more efficient surrogates or suitable umbrella species (Shi et al. 2019). For instance, a prioritization analysis by Di Minin et al. (2016) brought new insights into the discussion of the biodiversity surrogate power of large carnivores. Based on the analysis of extensive data, large carnivores would be better suited as surrogates for certain taxa, such as other mammals and amphibians, but probably not for reptiles and birds. However, this is probably not a panacea, because other GIS-based analyses using extensive datasets have also led to mixed observations of biodiversity surrogacy power (Shi et al. 2019; Escalante et al. 2020), which resonates with the findings of our overview of reviews on surrogacy literature.

Another promising avenue is to take advantage of conservation planning methods directly in identifying surrogates, rather than post hoc for applying them to practice. Ward et al. (2019) showed that there was a 7‐fold increase in management efficiency (threat abatement) when species on the Australian federal government’s umbrella prioritization list were replaced with others chosen with spatial optimization tools. If, instead, the starting point is that charismatic flagships will continue to be used in fundraising, then their effectiveness as biodiversity surrogates can be improved by using them in combination with other, well-surveyed taxa and habitat types (Di Minin and Moilanen 2014). An alternative is to use the representation of flagship species as a precondition in conservation planning primarily based on other data: McGowan et al. (2020) found that constraining site selection to flagship habitat caused a loss of 11–21% in species representation as compared with an optimal solution, implying that an equally large increase in fundraising capability would already compensate for the inefficiency. All of this suggests that surrogate literature should move from simple congruence assessments of individual taxa to more integrative approaches.

A taxonomic bias toward mammals and birds was evident in WTP research, but it aligns relatively well with what has typically been used for fundraising (Clucas et al. 2008; Smith et al. 2012), therefore providing us with relevant data for addressing differences between different flagship types. Similarly, the geographical bias toward developed countries is not a very severe issue as ENGOs have typically obtained most of their funding from the developed world (Charities aid foundation 2018). Nonetheless, more balanced sampling in future research would certainly help improve fundraising effectiveness, especially since growing economies will have a greater role in the field in the future.

WTP studies covered by the reviews we overviewed are highly heterogeneous, and we do not know how well they reflect actual donations to ENGOs. Some of the primary literature on WTP comes from the context of ecosystem service valuation and framing the questions as ecosystem service instead of donating to ENGOs could influence WTP in unknown ways. WTP studies are often conducted with the purpose of quantifying public support to inform choices of resource allocation for various policies. In this context monetary valuation has been seen as inappropriate to guide choices that are essentially moral or ethical (Spash 1997). It is also not entirely clear how well surveys formulated to address public spending reflect WTP as donation to ENGOs, even though typically tax as the payment vehicle resulted in higher WTP than voluntary payments when both kinds were included in meta-analysis (Lindhjem and Tuan 2012, although the opposite was found by Lindhjem 2006; Hjerpe et al. 2015; Amuakwa-Mensah et al. 2018). Even when WTP studies do address fundraising, they can be criticized for not reflecting real behavior due to hypothetical bias (List and Gallet 2001) and social acceptability bias (Börger 2012), and the WTP-estimates may be even 3-fold higher in hypothetical situations (List and Gallet 2001). But as we focus on relative differences between flagship types, the problem is less essential than it would be if the aim was to determine actual monetary value for the goods. In addition, BT could be very useful in conservation research, but its meaningful application would require larger amounts of better standardized primary research and research syntheses. Some of the meta-analyses we assessed had not been made with benefit transfer or comparison with other meta-analyses in mind, but as advised in the guidelines for meta-analyses by Nelson and Kennedy (2009), this should be considered and explicitly stated in future studies.

Similarly, methods used to assess biodiversity surrogacy power varied between meta-analyses, making comparisons between studies difficult. Even though much attention has lately been given to standardizing conservation research (Sutherland et al. 2004), the challenge of biased evidence remains (Christie et al. 2020). Our work too was fraught with difficulties due to lack of standard practice in the primary research and consequently in the meta-analyses as well. This is somewhat typical in the realm of conservation research, where the studied phenomena are complex, often multidisciplinary, and large amounts of homogeneous data for robust syntheses may be difficult to obtain (Sutherland et al. 2004). For example, Favreau et al. (2006) concluded that the methods and the contexts of the studies they reviewed were too heterogeneous to be able to conclude which surrogate types (e.g., indicator, keystone, or umbrella) are successful in terrestrial ecosystems. Varying definitions of surrogate types may also affect the results of literature searches performed for meta-analyses or overview of reviews. As a result, some relevant reviews may not have been found in database searches if they used terms in a different way or do not even mention any surrogate type. For example, ecological characteristics that make a species a keystone may have been studied in a paper that never mentions the term ‘’keystone species’’, leaving it out of search engine results.

Another issue to consider in future studies is the impact of human-wildlife conflicts especially on rural economically weaker households and businesses, leading to regional differences in attitudes toward different flagships (Douglas and Veríssimo 2013). Among the meta-regressions we reviewed, Martín-López et al. (2008) found lower WTP for species with any kind of negative economic impact, but further research is needed to disentangle this complex phenomenon. Using conflict species as conservation flagships could also be counterproductive regarding attitudes toward conservation (Linnell et al. 2000; Douglas and Veríssimo 2013). Large carnivores and elephants are often used as flagships despite their role in social conflict, but, on the other hand, they are often expected to have an indicator or umbrella role (Fernando et al. 2005; Maheshwari et al. 2014). While they may be effective in fundraising from e.g., tourists (Estifanos et al. 2018), the attitudes of tourists or even conservation practitioners towards conflict-related species may differ from those of locals (Kanagavel et al. 2014). Even among the local people, there may be differences in attitudes towards these species: attitudes towards lions have been more negative than attitudes towards elephants, although both belong to damage-causing species (Kanagavel et al. 2014; Sibanda et al. 2020). Some species associated with human-wildlife conflicts may also be suitable as flagships locally in areas where there is little or no human-wildlife conflict (Mekonnen et al. 2022). It is worth noting that human-wildlife conflicts are not restricted to developing countries (Garshelis et al. 2020).

Research is lacking on WTP to pure marketing flagships as opposed to conservation of the flagship itself: Does a charismatic figurehead increase WTP toward any conservation project, or do people prefer charismatic flagships only when the species itself is supported? An additional factor that would need to be considered to achieve a more comprehensive picture of the cost-effectiveness of various flagships as surrogates, is the variation in expenses of collecting different types of data (Mandelik et al. 2010). Charismatic flagships are usually well-known and conspicuous species with already existing distribution datasets and monitoring programs, which could favor their use as conservation tools, while at the other extreme there would be species that are poorly known and hard to detect. But holistic concepts are not directly comparable from this perspective, as their underlying idea is that allocating conservation actions is not constrained to a specific target or type of data.

Future WTP-studies would do well to also include measures of pro-environmental attitudes and values, such as biospheric values, as covariates (de Groot and Thørgesen 2012) that would increase our understanding of donation behavior and improve international meta-analyses and benefit transfer (Ready and Navrud 2006). As an additional advantage, it would produce information that could be used for donor segmentation. The role of knowledge about the flagships has also not been adequately dealt with. Future WTP studies should either provide information on the conservation status of the targets to the respondents (as recommended also by Martín-López et al. 2008), or measure the pre-existing awareness of the respondents, to reach a better understanding of its impact. Knowledge about the flagships in general, for example, their ecological role, has received even less attention. Another interesting avenue would be to look into the increasing concern for climate change, which could be synergistically taken advantage of in fundraising efforts. Quite surprisingly, we found no trace of such overviews, as apparently it has not been studied enough yet to be addressed in meta-analyses. In addition, future overviews could explore the role of charismatic species in raising conservation awareness, which was not assessed in this study. Other unexplored underlying factors that may influence environmental philanthropic behavior are the impact of ENGOs image as well as the success and length of the past campaigns that could be addressed in future studies. Further, because charismatic species have often received disproportionate conservation attention, as a consequence, there are prominent case examples of conservation success that have value in communicating positive messages to the public.

A key message to ENGOs based on our results is the need to diversify strategies in fundraising, as there is no evidence in favor of a particular strategy, but ideally fundraising considers different donor preferences and values. Fundraising campaigns may have limited possibilities for informing the potential donors at the time of fundraising but channeling some funds to increasing awareness of the plight of many uncharismatic species or about holistic targets could return via increased donations adding to the many benefits of raising awareness. Nonetheless, charisma could still be useful in marketing contexts where an image must make an instant impact, such as bus shelter advertising, but should be in a much smaller role when potential donors have the possibility to make informed choices between flagships based on prior or provided knowledge.

We found varying sensitivity to scope in the WTP meta-analyses. ENGOs could probably attract more funds if they were able to convince the donors that their projects have a broader scope and higher effectiveness. Quantifying conservation effectiveness is certainly not straightforward in research, nor in a donation situation where it may be quite ambiguous what will be done with the money. Nonetheless, it seems plausible that ENGOs could benefit from informing the donors about what can be achieved with the donations, especially among donor segments representing ‘effective altruism’. Yet there may be other segments for which standard economic theory could simply describe poorly the behaviors related to philanthropy, such as, ‘impact philanthropists’, who may prefer prominent, individual species, specific, delimited locations, or even adoption programs for animal individuals, where it is possible to observe the impact their donations are making.

It appeared that WTP was lower for restoration than for preservation purposes. Fundraising campaigns for restoration need to put effort into convincing potential donors of its effectiveness and necessity and explaining how it is a fundamental part in preventing extinctions and further losses. WTP appeared lower also for some less prominent terrestrial ecosystems as opposed to e.g., tropical, and marine ecosystems. Such targets that suffer from lower WTP, could perhaps benefit most from well-planned, strategic use of charismatic flagship species. Research is still needed to improve our understanding on how to reach different donor segments, but targeted marketing in social media has great potential for this purpose.