Biodiversity Offset Schemes for Indonesia: Pro et Contra

Abstract

Global biodiversity is in crisis, with wildlife populations declining 69% since 1970 (WWF). Preserving and restoring ecosystems is essential for sustaining life on Earth. However, many countries rely on market-based instruments like biodiversity offsets, despite little evidence of their effectiveness. This study critically examines biodiversity offsets, identifying institutional, data, ecological, economic, and social failures that undermine their success. Using Indonesia, a global biodiversity hotspot, as a case study, we develop an econometric model to analyze key drivers of deforestation. The findings reveal that biodiversity offset schemes are fundamentally flawed: they lack scientific credibility, rely on arbitrary ratios, lack auditing and transparency, create value conflicts, and fail to achieve “No Net Loss” even over a 100-year timeframe. Offsets do not compensate for lost biodiversity, especially for affected communities, and are rarely supported by ecosystem mapping or robust valuation metrics. Without major reforms, they cannot halt or reverse biodiversity loss. A stronger, evidence-based approach is urgently needed. Rather than relying on ineffective offset schemes, the global community must prioritize genuine ecosystem restoration and sustainable conservation strategies to protect biodiversity for future generations.

1. Introduction

Global biodiversity is experiencing a rapid decline, with species extinction rates now tens to hundreds of times higher than natural background levels, driven by land-use change, exploitation, climate change, pollution, and invasive species (IPBES, 2019) [1]. Despite the urgent need for systemic solutions, the global policy response has increasingly favored market-based instruments, particularly biodiversity offsets. These mechanisms, intended to compensate for ecological destruction by ensuring “No Net Loss” (NNL) or even a net gain of biodiversity, have gained widespread adoption despite limited empirical evidence of their effectiveness. Indonesia is one of the most biodiverse countries on the planet and is currently occupying a second spot in the Global Biodiversity Ranking (Brondízio et al., 2019) [2]. In 2002 the Delegates of Indonesia and other megadiverse countries and the Environment Ministers set up a Group of Like-Minded Megadiverse Countries (LMMC) to promote consultation and cooperation in Cancun, Mexico. Indonesia is a signatory to the UN Convention on Biological Diversity (UN, 1992) [3], which automatically makes UN CBD Kunming Montreal Biodiversity Framework (UN CBD, 2022) [4] adopted in 2022 international law that has to be adhered to. The UN CBD Kunming Montreal Biodiversity Framework responds directly to the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES) Global Assessment Report on Biodiversity and Ecosystem Services (IPBES, 2019) [1] to ‘catalyse, enable and galvanise urgent and transformative action by Governments’ to halt and reverse biodiversity loss. The Kunming Montreal Biodiversity Framework replaced the Aichi Targets for Biodiversity (Anonymous, 2020) [5], none of which have sadly been met, and recognizes that reversing biodiversity loss is the common concern of humankind, that the implementation of the framework should be based on scientific evidence and on the ecosystem approach of the CBD, and places biodiversity at the heart of sustainable development agenda (UN CBD, 2022) [4]. Biodiversity offsets are defined as “measurable conservation outcomes resulting from actions designed to compensate for significant residual adverse biodiversity impacts arising from project development and persisting after appropriate prevention and mitigation measures have been implemented. The goal of biodiversity offsets is to achieve no net loss, or preferably a net gain, of biodiversity on the ground” (BBOP, 2012) [6]. From purely provisioning flows of ecosystem services, regulation and maintenance ecosystem services, and cultural ecosystem services, ecosystems are fundamental to our survival. At the same time, nature is incredibly complex and non-linear. Many authors agree that the recent COVID-19 pandemic that has resulted in nearly 7 million deaths worldwide, has been ultimately caused by humans encroaching upon the ecosystems (Shmelev et al., 2023) [7]. We need, therefore, to apply a precautionary principle, which is enshrined within UN Convention on Biological Diversity (UN, 1992) [3]. There is a growing interest in assessing the true progress of countries towards sustainability and establishing the role that business could play in assessing and mitigating its impacts on nature and ecosystems (TNFD, 2023; Shmelev, 2018; Shmelev and Gilardi, 2025a, 2024b) [8,9,10,11]. These developments signal a recognition that conventional economic metrics are insufficient in capturing ecological reality or long-term viability. Building on the theory of ecological economics, which firmly positions the economic system within the wider environmental system and emphasizes our fundamental dependence on ecosystems (Shmelev, 2012) [12], this study aims to review the cutting-edge international experience in applying a policy tool of biodiversity offsets and shed new light on the current state of affairs with deforestation in Indonesia. The following Section 2 introduces a comprehensive literature review and builds a taxonomy of factors affecting the performance of biodiversity offsets. Section 3 provides an evidence-based analysis of factors affecting the biodiversity in Indonesia. Section 4 offers a comprehensive econometric model of deforestation in Indonesia. Section 5 offers a detailed discussion of the results and their policy relevance. Section 6 concludes.

2. Literature Review

We have carried out a comprehensive meta-analysis review of peer-reviewed literature available through international bibliographic databases inspired by the PRISMA methodology (Page et al., 2021) [13]. Based on the analysis carried out in the Scopus database, and using the keywords ‘biodiversity’ coupled with (‘offset’ or ‘mitigation banking’), we identified the key countries where the research has been focused. The body of literature focuses predominantly on Australia, Brazil, China, and the USA (Figure 1). Examining the literature by biome, we found that wetland and grassland biomes, as well as marine ecosystems, attracted the attention of most researchers (Figure 2).

Despite the ambitious global targets on biodiversity at the global scale, none of the Aichi targets have been met (Anonymous, 2020) [5]. This clearly indicates that something is not working in the global system of nature conservation, including biodiversity offsets and gives us a warning signal, encouraging us to investigate the causes of such ineffectiveness through an evidence-based assessment.

According to PRISMA-inspired analysis of peer-reviewed sources, there are serious grounds to believe that biodiversity offsets, as they are practiced currently around the world, are not achieving the desired effect (Figure 3). We will therefore aim to offer a way forward for Indonesia that could learn from the experience in different countries around the world.

The most comprehensive survey of biodiversity offsets so far attempted includes a study (Bull & Strange, 2018) [36] that cites 12,983 offset projects around the world currently in operation, covering 153,679 km² in 37 countries. The authors mention that the majority of the projects surveyed were focused on forests (66.7%) and wetlands (17.5%) and were located in boreal, Mediterranean, temperate and tropical forest biomes (92%) and 7% in grasslands.

Vatn (Vatn, 2015) [37] characterizes biodiversity offsets alongside payments for ecosystem services and carbon trading as part of the group of market-based tools aimed at handling environmental problems. The author provides a comprehensive framework to assess the constituent elements of constructed markets and shows that in many instances these are not ‘markets’ at all.

There are currently very intense debates regarding the effectiveness, the necessary conditions and potential implications of these schemes as outlined in, e.g., (Muradian et al., 2013) [38], who cite the example of Indonesia in the context of palm oil plantations having a negative impact on biodiversity and voluntary carbon markets not capable of acknowledging the benefits of forest as competitive with the price that could be obtained through selling palm oil on the market.

Our detailed meta-analysis presented in a large table in the annex is well-summarized in a structural diagram that identifies five key aspects of biodiversity offsetting: data, institutions, ecology, economics and society. With further detail provided in the accompanying tables, the diagram summarizes the key aspects discussed in peer-reviewed literature on the subject.

2.1. Institutions

Among the institutional aspects, the following issues have been flagged: the fact that biodiversity offsets are poorly regulated (Abdo et al., 2021) [14], not monitored (Lindenmayer et al., 2017) [32], not transparent (Maron et al., 2016) [15], rarely involve any auditing (Lodhia et al., 2018) [16], unethical and open to misapplication (Gordon et al., 2015) [17]; the researchers cite that 25% of projects did not use a management plan (Bezombes et al., 2019) [18]; offsets could be required to last for as long as the impacts of development, or in perpetuity, which is not forever but 50–75 years (Bull et al., 2013) [19]; offsets present value conflicts between environmental protection and economic development with information asymmetries abounding (Evans, 2023) [20]. In Australia, administrators use a “backloading” strategy that diminishes accountability and transparency and facilitates the use of offsets as the default option, deliberate blurring of environmental objectives to prioritize economic or social aims and in general, according the Australian government, there is a lack of evidence that offsets are effective and actually achieve their intended outcomes (Evans, 2023) [20].

In an article focused on Brazil, another megadiverse country, entitled ‘Playing Musical Chairs with Land Use Obligations’ (Filoche, 2017) [39] refers to Brazil’s historical refusal “to agree to Northern states using REDD to offset their polluting activity by purchasing carbon credits derived from non-deforestation in Southern countries”. This research quotes Sistema Nacional de Unidades de Conservacão—Conservation Units National System established as law in 2000 in Brazil. The projects making a significant environmental impact were required to pay a minimum of 0.5% of the project value into the special fund. The long legal battle involved going through various courts, aiming to limit the rate to a maximum of 0.5%, which still seems to be ongoing.

The most recent study by (Evans, 2023) [20] used semi-structured interviews surveying members of staff of the Department of the Environment and Energy in Australia and found that biodiversity offsets in Australia are very rarely effective: so called backloading takes place where regulators postpone the offset discussion until after the destruction of biodiversity is approved, the requirements are watered down in the negotiation process, verification of information is too time consuming and the government itself concluded that the scheme is not effective and is not achieving the stated objectives.

2.2. Data

Regarding the data-related aspects of the puzzle of biodiversity offsets, the researchers state that: there exists no single metric for the value of biodiversity (Bull et al., 2013) [19], the offset requirements are poorly defined (Abdo et al., 2021) [14], and limited or no measurement of environmental outcomes is carried out (Abdo et al., 2021) [14]. The researchers point out that the use of multiple metrics may result in a more comprehensive understanding of biodiversity losses and gains (Bull et al., 2013) [19], the baseline for defining requirements for demonstrating no net loss of biodiversity needs to be carefully chosen, and uncertainty in the offsetting outcome is dealt with by increasing compensation ratios (Bull et al., 2013) [19]. At the same time, the capacity to access this information by regulators was limited by short deadlines, physical separation between branches, and the loss of institutional knowledge through voluntary redundancy rounds (Evans, 2023) [20].

Research by (Bull et al., 2013) [19] stresses that biodiversity is not a fungible asset and cannot really be traded as a commodity. At the same time, the longevity issue is raised given the fact that perpetuity rarely means ‘forever’ and amounts to 50–75 years at most, which could be a ‘license to kill’ in relation to ancient and diverse ecosystems. Reviewing the experience of wetland banking in the US, (Maron et al., 2012) [40] states that although some ecological indicators, namely biomass and species richness often recover well in wetlands, others like soil physical and chemical properties, species composition, and nutrient cycling could take much longer to restore. Reviewing the 30 years of Species Conservation Banking (SCB) in the USA, (Carreras Gamarra & Toombs, 2017) [41] mention that 50% of all SCBs are located in California and cite that the “majority of SCBs do not include measures of habitat quality (79%) for credit calculation, and in 70% of banks, one acre (0.4 ha) equals one credit, meaning that credits are based only on the habitat area, irrespective of quality, with 6% using a mitigation ratio or multiplier for calculating credits, which is really far from being scientific and could hardly be applied for ancient, biologically diverse and incredibly valuable ecosystems in Indonesia.

2.3. Ecology

Examining the ecological dimensions, researchers raised concerns that biodiversity offsets are almost never supported by ecosystems mapping (Jacob et al., 2016) [21], that violations of additionality are omnipresent (Narain & Maron, 2018) [22], No Net Loss (NNL) is unlikely to be achieved for 146 years (Gibbons et al., 2018) [23], NNL is almost never observed in biodiversity offsetting schemes (Bigard et al., 2017) [24], biodiversity offsetting practice is not guided by scientific principles (Niner et al., 2021) [25], biodiversity offset was different from biodiversity lost (Hubert Ta & Campbell, 2023) [26], it is not clear if the NNL goal was achieved in Florida biodiversity offsetting (Levrel et al., 2017) [27], that like-for like goal of equivalence is difficult to reach for coastal offset projects (Stone et al., 2019) [28], that connectivity assessment is essential but is rarely performed (Sales Rosa et al., 2023) [29], the fact that biodiversity is not fungible calls into question the use of out of kind offsets (Bull et al., 2013) [19] and that in an empirical analysis in the context of Indonesia, priority areas for carbon and biodiversity offsetting hardly overlapped (Budiharta et al., 2018) [30].

New research by (Bigard et al., 2017) [24] assessed the way biodiversity requirements are taken into account in the context of Environmental Impact Assessment in the French context and stated that, in practice, No Net Loss (NNL) has almost never been observed, and often, the actual deterioration of biodiversity took place. The researchers state that “common reliance on offsetting to achieve NNL has received serious criticism due to the fact that offsets are rarely adequate, complete offsetting may be illusory due to the complexity of ecological processes and weak institutional organisation of the mitigation hierarchy impairs attempts to achieve NNL”.

The paper by (zu Ermgassen et al., 2019) [42] assessed 26 biodiversity offsets from 10 studies, of which only 9 achieved NNL for all given outcome variables, 7 failed to achieve any at all, and 8 achieved NNL for some outcome variables only but not for others. The authors conclude that historically NNL has been more successful in wetland than forested ecosystems and state that there exists a considerable gap “between the global implementation of NNL and the evidence base concerning ecological effectiveness”. The research showed that “67% of the world’s offsets are applied in forested ecosystems, yet our review reveals that only four studies have assessed NNL outcomes from offsets applied to forest ecosystems or wildlife. Of these, none demonstrated that their associated NNL targets were achieved” (zu Ermgassen et al., 2019) [42].

Reviewing the 30 years of Species Conservation Banking (SCB) in the USA, (Carreras Gamarra & Toombs, 2017) [41] mention that 50% of all SCBs are located in California and cite that the “majority of SCBs do not include measures of habitat quality (79%) for credit calculation and in 70% of banks one acre (0.4 ha) equals one credit, meaning that credits are based only on the habitat area, irrespective of quality. Six percent use a mitigation ratio or multiplier for calculating credits, and 4% determine credits based on the number of individuals of the covered species that the area could potentially accommodate”, which is really far from being scientific and could hardly be applied for ancient, biologically diverse and incredibly valuable ecosystems in Indonesia.

Ecosystem services must play a central role in biodiversity offsets (Jacob et al., 2016) [21], the position we take in this report. The researchers concluded that “biodiversity offsetting rarely considers human populations who suffer from environmental losses generated by development projects and those that benefit from offset actions, regardless of the level of dependency of local communities on ecosystem services in maintaining their livelihood. Including proposals may help to link human activities and amenities to affected or restored ecosystems, making the offsets more fair and ethical (Jacob et al., 2016) [21].

The research by (Budiharta et al., 2018) [30] focused on the island of Borneo in Indonesia and states that until recently, most offsetting studies focused on single impacts, most commonly biodiversity, as opposed to whole ecosystems with a complete spectrum of ecosystem services. The research found that priority areas for carbon and biodiversity offsetting in Borneo hardly overlapped at all due to the uneven distribution of ecosystem services.

In the paper with a telling title ‘Faustian bargains?’, (Maron et al., 2012) [40] reviews the experience of wetland banking in the US and states that although some ecological indicators, namely biomass and species richness often recover well in wetlands, others like soil physical and chemical properties, species composition, and nutrient cycling could take much longer to restore with some ecosystem functions taking decades to be restored to pre-disturbance state.

2.4. Economics

On economic issues, it was highlighted that biodiversity offsets present a clearly neoliberal policy (Apostolopoulou, 2016) [31], that the policy is controversial (Narain & Maron, 2018) [22], that financial equivalency for marine biodiversity is problematic because there is no agreement on how to value biodiversity (Niner et al., 2021) [25], that sophisticated tools for decision-making are useful in managing policies related to biodiversity offsets, but they do not resolve the fundamental.

Some views are openly critical of biodiversity offsets. Thus, (Apostolopoulou, 2016) [31] identifies biodiversity offsets as “a paradigmatic neoliberal policy, aiming at a further privatization, commodification, and financialization of non-human nature”. Moreover, the authors state that in the case of the UK, “biodiversity offsetting was considered as capable of creating various business opportunities for consultants, conservation banking companies, and brokers” (Apostolopoulou, 2016) [31]. In addition to this the authors specify that “the reasons for resisting ecosystem degradation are increasingly defined in terms of profitability, manifesting not only an ideological victory for capitalism but the creation of novel spaces for its operations by potentially opening new domains for capital accumulation”. A financial interest of local governments, whose finances are now not in the best of shapes has been also identified through the interview process: “in the face of decreasing public budgets and increasing competition many local and regional administrations across England have been involved in biodiversity offsetting with the aim to gain profits, a typical manifestation of the way the rescaling of governance promotes the further entrepreneurial character of rural and urban places”. The authors go as far as to state that “offsetting in fact establishes a new policy frame that creates socially and spatially uneven outcomes … and in this way offsetting is inextricably linked to questions of domination and uneven access” (Apostolopoulou, 2016) [31].

The literature we have analyzed clearly presents the plethora of arguments characterizing the ability of biodiversity offsets to compensate for the development impacts adequately as questionable (Abdo et al., 2021) [14]. The authors state that offsets are ineffective and have poor success rates if three crucial issues are not comprehensively addressed: (1) compliance and enforceability, (2) measuring environmental outcomes and (3) uncertainty and transparency (Abdo et al., 2021) [14]. The issues emerging are largely related to the governance and administrative failures (Abdo et al., 2021) [14]. Transparency is a hugely important factor in the context of biodiversity offsets (Maron et al., 2015) [15]. Lack of transparency may lead to discretionary application of rules, resulting in inequity among developers. Monitoring (Lindenmayer et al., 2017) [32] and auditing (Lodhia et al., 2018) [16] of the offsets during and after implementation are considered to be fundamental.

In the context of India, another megadiverse country, the biodiversity offsetting is regulated by the India Forest Conservation Act of 1980 with additional bodies like Compensatory Afforestation Fund (CAF) and the Compensatory Funds Management and Planning Authority recommended to be established by the court ruling in 2002 (Narain & Maron, 2018) [22]. The authors refer to the fact that the Indian government failed to establish an effective institutional mechanism and accumulated USD 5.7 billion of compensation funds while deforestation continued to take place. As a result, the Compensatory Afforestation Fund Act was passed in 2016. The government allowed the conversion of diverse natural forests for afforestation by monoculture plantation or carried out afforestation on lands belonging to indigenous tribes. The researchers also cite examples where project compensation funds have been utilized to fund already existing government commitments, thereby diminishing the net ecological impact of the programme.

Some researchers stated that financial equivalency for marine biodiversity is problematic because there is no agreement on how to value biodiversity (Niner et al., 2021) [25], that sophisticated tools for decision-making are useful in managing policies related to biodiversity offsets, but they do not resolve the fundamental conflicts of values that exist in politics and administration (Evans, 2023) [20], that the impossibility of defining a consistent, fungible unit that comprehensively captures biodiversity means that biodiversity itself is not a tradable market commodity (Bull et al., 2013) [19], and that the application of discount rates could lead to a disaster if future irreplaceable and catastrophic impacts of biodiversity and ecosystem service loss could be discounted. This position fully corresponds to the argument that a multi-criteria framework—and not the financial assessment framework used by Costanza et al. (2014) [43]—should be the guiding principle for the assessment of the true value of biodiversity, which I have argued since 2008 when I was a consultant to IUCN and 2009 when I was a consultant for UNEP TEEB.

2.5. Society

Social aspects of biodiversity assessment schemes have often been reported as amounting to the removal of nature from people (Kalliolevo et al., 2021) [33]; benefits do not compensate costs (Bidaud et al., 2018) [34]; offsets cannot compensate for the impacts on these communities that benefitted from water-related ecosystem services (Souza et al., 2021) [35]; greenness is associated with an increase in physical activity, positive mental health, reduced stress levels, lower incidence of allergies, reduced obesity, increased cognitive development of children but that is almost never taken into the account in administering biodiversity offsets (Kalliolevo, 2021) [33].

Research by (Bidaud et al., 2018) [34] focused on the assessment of the biodiversity offset schemes in the context of mining development in Madagascar, a low-income developing country. It concluded that “although it acknowledged the livelihood dependence of local people on natural resources and provided micro-development projects to support alternative livelihoods, Ambatovy’s biodiversity offset programme faced critical social issues. Firstly, though acknowledging the positive impact of some of the development projects on their lives, local stakeholders felt that they had suffered a net cost from the biodiversity offset, as the benefits from the alternative livelihood activities did not compensate for the costs of the conservation restrictions. Secondly, those who benefited most from the development projects were neither those who bore the greatest costs of forest access restriction nor the poorest people but tended to be those with more power locally (Bidaud et al., 2018) [34].

3. Analysis of the State of the Environment in Indonesia

Indonesia has come a long way in terms of development and managed to achieve a great deal in the past several decades. At the same time, there are worrying trends related to the state of the environment in Indonesia that we would like to highlight. We will treat forest cover as a first step in understanding the state of biodiversity in Indonesia. Deforestation in Indonesia continues to happen (Boentoro & Wherrett, 2021) [44], (Jati et al., 2018) [45]. In fact, between 1990 and 2021, the forest cover in Indonesia was reduced from 65% to less than 50% of the total area.

3.1. Mining

Coal production in Indonesia is increasing (Figure 4), affecting the share of forested land. Mining is a factor affecting biodiversity in Indonesia, given that half of the world’s mining-related biodiversity loss occurred in Indonesia, Australia and New Caledonia (Cabernard & Pfister, 2022) [46]. This happens while major trade flows associated with biodiversity loss are related to sales of Indonesian coal to India and China. Exports of coal represent 11.5% of Indonesia’s exports and amount to a figure of USD 28.4 billion. Most of the export of coal from Indonesia finds its way to China (31.1%), India (16.1%), Japan (9.49%), the Philippines (8.66%), Malaysia 8.95%), and South Korea (7.35%).

Research by (Cabernard & Pfister, 2022) [46] shows in a high amount of detail how the mining of coal and other minerals in Indonesia is contributing to the production activities in China and India, which is in-turn, resulting in the creation of infrastructure for construction, electronics, machinery and transport.

Indonesia has become the largest nickel producer in the world, with rates of nickel mining increasing substantially since the 1990s (Figure 5). The amount of available forest tends to be negatively correlated with the amount of nickel mined in Indonesia. Indonesia produces around 37% of the global supply of nickel and holds approximately 22% of global reserves. Around USD 14 billion has been invested in nickel smelting in Indonesia. Between 2022 and 2029, it is likely that Indonesia’s nickel production will reach the level of 75% of the global supply. The average nickel price in 2021 reached USD 17,489 as opposed to USD 13,787 per tonne in 2020. Multiple examples of post-mining regeneration of land exist, although this does not reverse the overall negative trends for biodiversity in Indonesia (Novianti et al., 2018) [47], (Woodbury et al., 2020) [48], (Isworo & Oetari, 2023) [49].

Copper, nickel and bauxite have also shown an increase in production, especially after the bans on export were reduced. Overall, the mining sector is one of Indonesia’s strongest contributors to the economy, especially post-pandemic when the prices for mineral, coal and gas resources increased. PwC also points out that mining companies are often the only employers in remote areas and that the demand for critical minerals used in energy transition will increase as part of the rush towards clean energy, while coal and other resources will decline (PwC, 2023) [50].

Multiple examples of post-mining regeneration of land exist, although this does not reverse the overall negative trends for biodiversity in Indonesia: (Novianti et al., 2018) [47], (Woodbury et al., 2020) [48], (Isworo & Oetari, 2023) [49].

3.2. Agriculture

Palm oil continues to be one of the main drivers of deforestation (Yu et al., 2023) [51]. As Figure 6 illustrates, the palm oil production in Indonesia massively increased. The export of palm oil in Indonesia amounted to USD 27.3 billion in 2021, making it the largest palm oil exporter in the world. It contributes a significant proportion of foreign trade revenues for Indonesia and accounts for 11% of its exports. Most of the palm oil export goes to China (15.5%), India (12.6%), Pakistan (10.4%), but also the USA (5.07%), Spain (3.61%), Russia (2.83%), Italy (2.47%), and the Netherlands (1.93%). Indonesia’s palm oil production gives employment to approximately 16 million workers. The presence of roads tends to intensify the expansion of smallholder palm oil plantations even further (Zhao et al., 2022) [52]. Despite the recent reduction in deforestation from palm oil expansion, there has been a recent increase in deforestation, as reported by (Jong, 2024) [53].

Rice production (Figure 7) tends to be inversely correlated with the amount of forest available, however, it does not seem to be the most significant factor of deforestation as the produce is mostly used for internal consumption. Significant quantities of rice are imported into Indonesia from India (30.4%), Singapore (18%), Malaysia (15.4%), Thailand (15.1%), and Vietnam (12.7%). Imports of rice tend to exceed exports by approximately USD 250 million.

Among various factors exerting pressure on ecosystems and biodiversity in Indonesia is the development of infrastructure (Figure 8). Serious concerns are being raised by scientists about the forthcoming developments in Borneo associated with the transfer of the capital of Indonesia to the island (Spencer et al., 2023) [54], with the proximity of roads identified as one of the crucial factors for the reductions in biodiversity.

Indonesia is the world’s largest producer of oil palm, producing over 30 million tonnes of palm oil annually (UNDP, 2024) [55]. The palm oil industry generates 4.5% of Indonesia’s GDP and offers employment opportunities to 3 million people (UNDP, 2024) [55]. The production satisfied 50% of the EU’s palm oil needs and 31% of the Dutch demand (Ministerie van Landbouw, 2023) [56]. The current demand trend will increase palm oil production by 12 million tonnes by 2035 (Smith, 2023) [57].

Indonesia has taken steps to create more sustainable palm oil production. One such step is the partnerships with the Netherlands to address unsustainable practices and reduce deforestation without having to compromise on production (Wageningen University, 2023) [58]. One of the programmes is named ‘SustainPalm’ and it lasts for three years (Ministerie van Landbouw, 2023) [56], while the other is the National Initiative on Sustainable and Climate Smart Oil Palm Smallholders (NI-SCOPS), aimed at providing the Indonesian government with support to meet its commitments under international agreements (IDH, 2024) [59].

Additionally, the Green Commodities Programme and Good Growth Partnership collaborated with the national government to establish the Indonesia Sustainable Palm Oil Platform (FoKSBI) as a neutral ground aimed at facilitating multi-stakeholder conversations on the challenges of sustainable palm oil development in Indonesia (UNDP, 2024) [55].

Furthermore, President Joko Widodo signed in 2019 the National Action Plan for Sustainable Palm Oil, encompassing six different production regions (UNDP, 2024) [55]. The Plan sets out to create better coordination between smallholders of palm oil farms and provide them with training and access to various resources, such as financing. It also highlights the importance of effective governance, from law enforcement to conflict resolution. Additionally, the plan aims to improve environmental management and monitoring through various measures, e.g., biodiversity conservation, reduction in GHG emissions, etc. The Indonesian Palm Oil (ISPO) standard also requires greater acceptance and a smoother certification process for smallholders under it (UNDP, 2014) [60]. The ISPO is regulated by Regulation of the President of the Republic of Indonesia No. 44 of 2020 on the Indonesian Sustainable Oil Plantation Certification System.

While sustainable palm oil practices appear to gain traction in Indonesian policies, their application differs in reality. Putri et al. pointed out that while ISPO certification would be beneficial for all parties involved (Putri et al., 2022) [61], the legislative process sequestrates these positive outcomes. The regulations and policies are produced at the central government level, which ignores all the subtleties of lower governance levels and makes transposition difficult. The process is even more complex when, at a lower level, there are regulatory guidelines that contradict each other when it comes to implementing sustainability certificates. Or, alternatively, there are no guidelines (Putri et al., 2022) [61].

3.3. Infrastructure

In May 2023, the Ministry of National Development Planning/National Development Planning Agency (BAPPENAS) published its Public Private Partnership Infrastructure Projects Plan in Indonesia (Ministry of National Development Planning, 2023) [62], outlining the status of future infrastructure plans. According to the regulation, there are two types of PPP project schemes—solicited, which are initiated by the Government, and unsolicited, which are initiated by members of the private sector. The document provides an outline of projects at different preparation stages where their realization aspects are considered, such as feasibility, budget, public consultation, etc. (Ministry of National Development Planning, 2023) [62]. In the PPP book, there is a list of 52 projects to be developed, out of which 17 projects relate to transportation and are outlined below.

3.4. Transport

The Public-Private Partnership Infrastructure Plan in Indonesia Book mentions the development of two airports—Singkawang and Bintan. The estimated financial cost of Singkawang is USD 117.00 million, while Bintan would cost USD 755.26 million. The Singkawang airport is intended to improve the accessibility of people and goods in the area, preventing economic paralysis due to transportation shortage. The relocation of the country’s capital from Jakarta is also emphasized as a reason for the airport. Comparatively, for Bintan, the airport is built for tourism and industrial activities (Ministry of National Development Planning, 2023) [62]. Additionally, the Baubau Port, located in the Southeast Sulawesi district, is planned to be expanded to facilitate the economy and tourism. The project is estimated to cost USD 16.95 million (Ministry of National Development Planning, 2023) [62].

The Bandung Railway project is a light rail transit aimed at improving urban traffic by making it faster and more reliable, costing approximately USD 785 million (Ministry of National Development Planning, 2023) [62]. There are also three type A bus terminal development plans located in different provinces to enhance transportation services with mixed-use terminal schemes providing residential, commercial, industrial and entertainment purposes to the terminals, operating on different floors of the complex. These new developments would all amount to approximately USD 111 million (Ministry of National Development Planning, 2023) [62].

The PPP Book also includes plans for eight toll roads and two bridges functioning on the toll road principle. The Demak-Tuban toll road project is the longest and most costly toll road project outlined in the PPP Book, reaching an estimated cost of USD 3440.27 million and stretching over 179.55 km. The idea behind the project is to complete an already existing toll road (Semarang-Demak) and to elevate the development of the areas surrounding toll roads as well as developing economic growth (Ministry of National Development Planning, 2023) [62]. The Batam-Bintan bridge will cost approximately USD 1029.60 million and exten over 7684 m. The objective is to increase trade and industries in the two islands, as well as take advantage of the geographical location of Batam near Singapore and Malaysia (Ministry of National Development Planning, 2023) [62]. None of the projects mentioned above mention the environmental impact or their environmental benefits. The general reasons behind creating them are of an economic and social nature, namely facilitating trade, tourism and interconnectedness.

The increasing urbanization and economic development of Indonesia require more transportation possibilities, which is reflected in the Partnership Plan. However, more roads mean more motor vehicle usage, which affects air quality. Indonesia has exceeded the World Health Organization’s (WHO) recommended range of PM2.5 since 2022, which affects public health (Ernyasih et al., 2023) [63]. In fact, in Batam city, where a bridge was developed for industrial and trade purposes, the PM2.5 levels were as high as 45 μg/m³, while carbon monoxide (CO) was recorded to be 35.83 ppm (Ernyasih et al., 2023) [63].

It is worth noting that the WHO recommends that the annual level of PM2.5 does not exceed 5 μg/m³ (WHO, 2021) [64]. South Tangerang, a city filled with industrial areas and characterized by rapid trade growth and highly crowded areas, has been deemed the most polluted city in Indonesia, as data found a positive correlation between the increase in human health disorders and transportation, industrial activities and air pollution (Listyarini et al., 2021) [65].

Heavy metal contamination around ports is also a currently debated issue in the Indonesian scientific community. The Belawan Harbor, one of Indonesia’s busiest ports, was found to be lightly polluted with cadmium and lead between November 2018 and January 2019 (Sulistyowati et al., 2023) [66]. If neglected, pollution could increase over the years due to human activity and lead to the extinction of endemic species, conflicts between the local community and companies, as well as a lack of jobs for fishermen. A similar result was found in the coastal area around the province of Daerah Istimewa Yogyakarta, where high levels of lead were found as a result of anthropogenic intervention—such as corrosion of metals, activities related to mining, painting of boats, etc. (Asih et al., 2022) [67].

Mixed-use terminal schemes, as part of transit-oriented development, are also a questionable option from a sustainability perspective. Specifically in Indonesia, especially in Jakarta, green open spaces saw a decrease in size due to the increase in public facilities (Hasibuan & Mulyani, 2022) [68].

Of 42 Strategic Priority Projects, most target infrastructure for roads, railways, new cities, and Special Economic Zones (SEZs). The rest cover the energy, mining, forestry (especially watershed recovery), marine fisheries, and manufacturing sectors. In general, the project descriptions would benefit from an enhanced green growth orientation and, without such review, could pose threats to natural capital and climate. The high-level policy goals for green growth thus must be further translated and mainstreamed into the design of such strategic projects.

The international scientific community is seriously concerned about the fact that ‘planned and ongoing road and rail line developments will have many detrimental ecological impacts including fragmenting large expanses of intact forest’ (Figure 8). More specifically, the authors are concerned that the landscape connectivity could decline from 89% to 55% if all the imminently planned projects go ahead (Figure A1). It is emphasized that the developments are likely to have significant impacts on rare species of rhinoceros, orangutans and elephants. It is underlined that planned infrastructure expansion will likely affect 42 protected areas, undermining Indonesian efforts to achieve key objectives of the CBD (Alamgir et al., 2019) [69].

The ground-breaking research by (Spencer et al., 2023) [54] revealed that up to 46% of critical habitat for threatened mammals is likely to be affected by the combined areas of road development and capital relocation. Proximity to roads has been identified as one of the most critical factors for the presence of mammals.

4. Multi-Factor Econometric Model of Ecosystem Deterioration in Indonesia

As one of the most biodiverse countries on the planet, Indonesia’s ecosystems are of global significance. However, given the massive deforestation that happened in the past decades (Figure 9), Indonesia cannot process the CO₂ emissions that its economy is producing. This dictates the need to establish firm scientific connections between various forms of economic activity and the nature and ecosystems in Indonesia.

To this end, we will attempt to create a comprehensive statistical model that will help explain the changes in the forest cover observed.

Based on the analysis of peer-reviewed literature and using a well-defined methodology, (S. E. Shmelev & Speck, 2018) [9] we identified several groups of factors (Figure 10), i.e., income, society, agriculture, mining, and infrastructure, that are responsible for the deforestation in Indonesia. More specifically, population growth has been found to negatively affect the available amount of forest in Indonesia, and GDP was shown to be a significant stabilizing factor in a multivariate regression model, which is presented in

Table 1. Statistical coefficients in the multi-factor regression model of deforestation in Indonesia, the estimation sample: 2000–2020.

	Coefficient	Std. Error	t-Value	t-Prob	Part. R2
Constant	73.6299	2.374	31.0	0.0000	0.9826
Population	−0.000000116	0.00000001.41	−8.20	0.0000	0.7981
GDP	0.00157451	0.0002517	6.26	0.0000	0.6972
Palm	−0.161774	0.01914	−8.45	0.0000	0.8078
Coal	−0.00633069	0.002666	−2.37	0.0296	0.2491
POM	−0.724765	0.2578	−2.81	0.0120	0.3173

. Agricultural production has been represented by palm oil production, which is reported to be one of the most serious factors of deforestation. Mining has been represented by coal mining, nickel mining and gold mining. Infrastructure has been represented by the overall volume of roads built in Indonesia.

A simple multi-factor econometric model (Figure 11) connecting the amount of forest in Indonesia with five drivers following a framework presented in Figure 5 clearly shows that all of them are connected with deforestation, albeit in different ways (Table 1). As the model output clearly shows, population growth is putting pressure on the available forest, and the production of palm oil and extraction of coal are statistically significantly correlated with the amount of forest available, with GDP levels in current international USD per capita improving the situation somewhat and acting as a stabilizer, while the palm oil moratorium introduced in 2011 by a decree, surprisingly, features a negative coefficient.

Building a model depicted in Table 1 and Figure 11 is a highly complex creative process that followed the methodological stage (Figure 10), analysis of the literature of the subsequent chapters in this report, analysis of data presented in Figure 5, Figure 6, Figure 7, Figure 8 and Figure 9, and multiple brainstorming sessions (pairwise regressions between the variables in the model are depicted in Appendix A).

Various additional statistical tests have been carried out with the help of OxMetrics 9.3 software to evaluate and extend the initial modeling approach. Our analysis began with a conventional linear regression model using coal production, palm oil output, GDP per capita, and population as predictors of forest cover change in Indonesia. This model demonstrated an adjusted R² of 98.93%, suggesting that over 99% of the variation in forest cover could be explained by the selected variables. All coefficients were statistically significant at the 5% level, with especially strong contributions from palm oil production, GDP, and coal extraction. Figure 11 illustrates the deforestation model output for Indonesia over the period 2000–2020. The palm oil variable, with the highest partial R², emerged as the single strongest contributor to forest loss, reflecting the economic and institutional prominence of the palm oil sector, which currently employs millions of Indonesians. GDP per capita, with a stabilizing effect, suggests that income growth may mitigate environmental degradation if aligned with value-added industrial strategies, such as Indonesia’s recent investment in lithium battery production. Coal production, by contrast, had a negative coefficient, underlining the destructive ecological impact of open-pit mining.

While this initial model was informative, deeper diagnostic testing revealed important limitations. The RESET test indicated potential model misspecification, and the condition number signaled extreme multicollinearity, especially between GDP, population, and aggregate commodity volumes.

Before re-estimating the model using log-log specifications, we conducted a correlation analysis among the independent variables. The raw-level variables—particularly GDP, population, coal, and palm oil—showed very high intercorrelations, suggesting multicollinearity. To address this issue, we recalculated all explanatory variables on a per capita basis. The resulting correlation matrix, presented in Figure 12, shows a marked reduction in correlations among key regressors, thus justifying the transformation and supporting the decision to pursue a nonlinear model specification.

In response, we pursued a transformation of the model to a log-log specification, using per capita expressions for all explanatory variables. This approach had both theoretical and statistical advantages. It reduced the condition number to below 3 × 10⁴, lowered the AIC from −2.72 to −9.34, and offered a more nuanced elasticity-based interpretation of relationships (

Table 2. Statistical coefficients in the multi-factor nonlinear regression model of deforestation in Indonesia, the estimation sample: 1990–2020.

	Coefficient	Std. Error	t-Value	t-Prob	Part. R2
Constant	10.0121	3.506	2.86	0.0082	0.2320
LOG_POP	−0.932304	0.3732	−2.50	0.0189	0.1877
LOG_COAL_POP	−0.049530	0.02411	−2.05	0.0497	0.1352
LOG_PALM_POP	0.108156	0.05284	2.05	0.0505	0.1314

). In this intermediate model, both population and extraction of coal per capita emerged as significant drivers, with palm oil appearing as significant at 10% albeit with a positive coefficient.

To capture further ecological and institutional dimensions, we extended the model to include additional per capita variables: nutmeg production, road infrastructure, coffee output, and gold extraction. This full nonlinear model (Figure 13) produced the strongest overall performance, with an adjusted R² of 98.76% and an AIC of –10.5367 (

Table 3. Statistical coefficients in the multi-factor nonlinear final regression model of deforestation in Indonesia, the estimation sample: 1990–2020.

	Coefficient	Std. Error	t-Value	t-Prob	Part. R2
Constant	12.2805	1.9630	6.26	0.0000	0.6298
LOG_POP	−1.19527	0.2064	−5.79	0.0000	0.5931
LOG_COAL_POP	−0.02601	0.0183	−1.42	0.1690	0.0806
LOG_PALM_POP	0.10349	0.0344	3.01	0.0063	0.2822
LOG_GOLD_POP	−0.02841	0.0594	−4.78	0.0001	0.4987
LOG_ROAD_POP	0.15526	0.0655	2.37	0.0266	0.1961
LOG_COFFEE_POP	−0.09504	0.0311	−3.05	0.0057	0.2882
LOG_NUTMEG_POP	−0.01989	0.0094	−2.10	0.0467	0.1611

). All key coefficients were statistically significant, apart from LOG_COAL_POP, and reasonably consistent with theoretical expectations. Importantly, residual diagnostics confirmed no evidence of autocorrelation, heteroskedasticity, but minor issues with non-normality. The role of nutmeg and coffee suggests that agroforestry dynamics beyond palm oil also shape land-use change, while road expansion continues to represent a critical vector for spatial transformation. The improved specification, lower multicollinearity, and enhanced residual behavior together justify this model as the most reliable framework for understanding the economic and environmental determinants of deforestation in Indonesia.

Furthermore, we tested two alternative models of deforestation based on the University of Maryland intact primary forest data [70], both in the linear form (

Table 4. Statistical coefficients in the multi-factor linear regression model of intact primary forest in Indonesia, the estimation sample: 1993–2017, UMD.

	Coefficient	Std. Error	t-Value	t-Prob	Part. R2
Constant	224.150	9.59	23.4	0.0000	0.9698
GOLD	−0.0191076	0.007158	−2.67	0.0162	0.2954
ROAD	0.0406448	0.01983	2.05	0.0562	0.1982
COAL_3	0.0134467	0.005605	2.40	0.0282	0.2529
COFFEE_3	0.0000152541	0.000003986	3.83	0.0013	0.4628
COCOA_-3	−0.00000772359	0.000002382	−3.24	0.0048	0.3822
POP_MLN	0.776153	0.06361	−12.2	0.0000	0.8975
PALM	0.327949	0.1078	3.04	0.0074	0.3523

) and the log form (

Table 5. Statistical coefficients in the multi-factor logarithmic regression model of intact primary forest in Indonesia, the estimation sample: 1993–2017, UMD.

	Coefficient	Std. Error	t-Value	t-Prob	Part. R2
Constant	42.3105	4.145	10.2	0.0000	0.8597
LOG_PPOP	−4.85222	0.4540	−10.7	0.0162	0.8705
LOG_ROAD_POP	0.174791	0.09562	1.83	0.0852	0.1643
LOG_GDP_CURR_INT	0.523938	0.07128	7.35	0.0000	0.7607
LOG_COAL_POP_-3	−0.0693475	0.03570	−1.94	0.0689	0.1816
LOG_COFFEE_POP_-3	−0.155286	0.05971	−2.60	0.0187	0.2846
LOG_PALM_POP_-3	0.109484	0.06213	1.76	0.0960	0.1545
LOG_GOLD_POP_2	−0.0170587	0.008597	−1.98	0.0636	0.1880
LOG_PALM_POP_1	−0.107663	0.05881	1.83	0.0848	0.1647

). The characteristics of residuals in these models are given in Figure 14 and Figure 15. We then produced a comparative table outlining the alternative models, accompanied by a diverse range of statistical characteristics of their performance (

Table 6. Comparative deforestation model diagnostics, Indonesia, 1990–2022.

Metric/Variable	Linear Model (WB)	Log Model (WB)	Linear Model (UMD)	Log Model (UMD)
Years	2000–2022	1990–2020	1993–2017	1992–2017
Observations	23	31	25	26
R2	0.9893	0.9876	0.9967	0.9988
Adj. R2	0.9862	0.9838	0.9954	0.9983
Sigma	0.2289	0.00462	0.6827	0.00651
AIC	−2.73	−10.54	−0.51	−9.8
AR 1–2	✅ F(2,15) = 0.53, p = 0.60	✅ F(2,21) = 0.14, p = 0.14	✅ F(2,15) = 0.81, p = 0.46	✅ F(2,15) = 0.67, p = 0.53
ARCH 1–1	✅ F(1,21) = 0.11, p = 0.75	✅ F(1,16) = 1.98, p = 0.18	✅ F(1,23) = 9.2 × 10−5, p = 0.99	❌ F(1,24) = 8.06, p = 0.009
Normality	✅ χ2(2) = 3.34, p = 0.188	❌ χ2(2) = 7.28, p = 0.026	✅ χ2(2) = 2.22, p = 0.33	❌ χ2(2) = 4.89, p = 0.009
Hetero-scedasticity	✅ F(9,13) = 0.61, p = 0.77	✅ F(14,16) = 1.66, p = 0.164	✅ F(14,10) = 0.96, p = 0.54	✅ F(16,9) = 2.18, p = 0.12
RESET	❌ F(2,15) = 11.92, p = 0.0008	❌ F(2,21) = 7.55, p = 0.0034	✅ F(2,15) = 3.65, p = 0.0509	✅ F(2,15) = 0.79, p = 0.4730
GDP	0.002 ***			0.524 ***
Population	−0.00000012 ***	−1.195 ***	0.776 ***	−4.852 *
Coal	−0.006 *	−0.026	0.013 *	−0.069 †
Road		0.155 *	0.041 †	0.175 †
Coffee		−0.095 **	0.0000153 ***	−0.155 *
Palm	−0.162 ***	0.103 **	0.328 †	0.109 †
Cocoa			−0.0000772 **
Gold		−0.028 ***	−0.019 *	−0.017 †
POM	−0.724 *
Nutmeg		−0.020 *

Notes: † p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001; ✅, significant results, ❌, insignificant results.

).

Palm oil exhibits contrasting effects across the four models, reflecting differences in specification, data source, and functional form. In Model 1 (linear, WB data), the coefficient is strongly negative (−0.162, p < 0.001), indicating that higher palm oil production is associated with forest cover loss—aligning with long-standing evidence of palm expansion into forested areas. However, this model fails the RESET test (p = 0.0008), suggesting potential omitted non-linearities or interactions.

In Model 2 (log-log, WB data), the relationship reverses direction: palm oil shows a positive elasticity (+0.103, p < 0.01). This result is counterintuitive, suggesting that higher palm oil output is associated with increased forest cover, which is inconsistent with most land-use change literature. The RESET failure (p = 0.0034) and non-normal residuals (p = 0.026) raise concerns about model misspecification, casting doubt on this finding. While Model 2 has the lowest AIC (−10.54) across all models, its diagnostic weaknesses warrant caution in interpreting this elasticity.

In contrast, Model 3 (linear, UMD data) finds a positive and statistically significant effect (+0.328, p = 0.0074). This may reflect palm expansion into areas already deforested before the time window, or increased yields without new land conversion (e.g., via intensification). Unlike Models 1 and 2, Model 3 passes all key diagnostics, including RESET (p = 0.0509), normality, and heteroscedasticity, making it structurally sound.

Finally, Model 4 (log-log, UMD data) reveals a negative coefficient (−0.107, † p < 0.10)—again suggesting a direct link between palm oil and deforestation, although at lower significance. This model passes the RESET test but fails normality (p = 0.009) and ARCH (p = 0.009), indicating potential volatility clustering in residuals. Its AIC (−9.8) is the lowest among UMD-based models, but not across all four.

The Palm Oil Moratorium (POM) appears only in Model 1 and has a large negative coefficient (−0.724, p = 0.012), implying that forest cover continued to decrease even after the moratorium’s 2011 implementation. This finding raises concerns about the short-term effectiveness of the policy, suggesting that enforcement gaps, legacy concessions, or delayed implementation may have undermined its intended goals.

Population has mixed effects across the models but is mostly negatively associated with forest cover. In Model 1 (Linear, WB), population shows a negative coefficient (−0.000000116, p < 0.001), meaning that higher population levels correlate with lower forest cover. In Model 2 (Log, WB), the elasticity is also negative (−1.20, p < 0.001), confirming that forest cover decreases proportionally with population increases. Model 4 (Log, UMD) follows this trend with a negative elasticity (−4.852, p < 0.05), indicating consistency across specifications when modeling forest cover. However, in Model 3 (Linear, UMD), population surprisingly shows a positive and statistically significant effect on forest loss (+0.776, p < 0.001). This suggests that areas with higher populations tend to experience less forest loss in hectare terms, a finding that may appear counterintuitive. It could reflect localized conservation, urban saturation, or data aggregation effects, particularly if population growth is concentrated in urbanized regions not undergoing active deforestation.

GDP is included only in Model 1 and Model 4, and its effects are consistently positive. In Model 1 (Linear, WB), GDP has a positive coefficient (+0.00157, p < 0.001), suggesting that economic growth is associated with increased forest cover. This may reflect improved environmental governance, afforestation efforts, or stronger conservation policies that accompany rising national income. In Model 4 (Log, UMD), the GDP elasticity is also strongly positive (+0.524, p < 0.001), meaning a 1% increase in GDP is associated with a 0.524% increase in forest cover. This reinforces the view that economic development, in certain contexts, can coincide with forest gains. However, GDP is absent from Models 2 and 3, so its effects on forest dynamics in those model structures remain unexamined.

Coal shows contrasting effects across models. In Model 1 (Linear, WB), coal production is negatively associated with forest cover (−0.00633, p = 0.03), indicating that increased coal output correlates with declining forest cover—a result consistent with expectations about mining and land degradation. In Model 2 (Log, WB), coal also has a negative elasticity (−0.026), but this effect is not statistically significant (p = 0.169), so no strong conclusion can be drawn. In contrast, Model 3 (Linear, UMD) finds a positive coefficient for coal (+0.0127, p = 0.028), implying that higher coal production is associated with less forest loss, a counterintuitive result that may reflect regional development dynamics or limitations in deforestation attribution. Model 4 (Log, UMD) again shows a negative elasticity (−0.069, p = 0.069), suggesting a tendency toward reduced forest cover with more coal production, though this is only marginally significant. These mixed findings highlight either model sensitivities or deeper structural variation in how coal impacts forest landscapes. Gold mining is a robust driver of deforestation in the models where it appears. In Model 2, it has a negative effect on forest cover (−0.028, p < 0.001), consistent with degradation from mining activities. In Model 3, lagged gold production is linked to greater forest loss (+4983 ha, p = 0.009), suggesting that the environmental impact emerges with a time delay due to permitting, infrastructure development, or post-extraction clearance.

Coffee cultivation presents a nuanced relationship with forest cover. In Model 2, coffee shows a negative elasticity (−0.095, p < 0.01), implying that increased coffee production is proportionally associated with forest cover loss, likely due to land conversion. A similar negative elasticity appears in Model 4 (−0.155, p < 0.05), reinforcing this deforestation link when considering proportional changes. However, Model 3 presents a contrasting picture: the coefficient is positive (+0.0000153, p < 0.001), indicating that higher levels of coffee production are associated with greater forest availability in absolute terms (hectares not lost). This suggests that in some contexts, coffee cultivation may coexist with forests, potentially due to agroforestry practices or regulations that limit clear-cutting. The divergence between the log-log models and the absolute model underscores the importance of scale and land-use context in interpreting the environmental impact of coffee production.

Nutmeg is only included in Model 2, where it shows a statistically significant negative effect (−0.020, p = 0.05), indicating that nutmeg cultivation contributes to forest cover decline.

Cocoa, included only in Model 3, has a significant lagged negative association (−2.92 ha, p = 0.018) with forest loss. This might be due to intensification strategies, sustainable certification, or a shift toward planting on degraded land rather than converting new forest areas.

Road infrastructure shows a consistently positive association with forest outcomes. In Model 2, road length has a positive elasticity (+0.155, p = 0.027), suggesting that increases in road infrastructure are proportionally associated with more forest cover, not less. This pattern holds in Model 3, where the coefficient is also positive (+54.49 ha, p < 0.01), indicating that regions with more road development tend to have greater forest area in absolute terms. Similarly, Model 4 confirms this with a positive elasticity (+0.175, p < 0.1). These findings may reflect the role of roads in improving land governance, accessibility for sustainable land use practices, or road-building policies that avoid core forest areas.

The analysis of the state of the art in forest cover and factors attributed to its reduction is hugely important in the context of the present article as it presents the background on which the future system of biodiversity offsets will have to be designed. Knowing the most important drivers of biodiversity loss supported by econometric evidence provides operational capabilities to the government of Indonesia in adapting and refining its strategic approach that focuses on managing biodiversity. It has to emphasize that the modeling presented in this chapter is unique and has been carried out for the first time ever.

5. Discussion

While the central government of Indonesia is actively taking steps to align with the international standards on green and sustainable growth, the existing policy landscape does not reflect that. On the contrary, it still provides opportunities to exploit resource-intensive development plans (Anderson et al., 2016) [71]. Promoting green, sustainable practices is also a divided arena, with some governmental bodies and NGOs taking initiative, while corporations and other actors who perceive the green economy as a constraint to economic development oppose them (Anderson et al., 2016) [71]. There is a lack of incentive structures and a solid regulatory basis to redirect stakeholders towards better, more sustainable practices (Brockhaus et al., 2012) [72]. ‘Decluttering’ the legal framework and simplifying it would also increase transparency, accountability and encourage coordination at different levels of governance (Brockhaus et al., 2012) [72].

At the same time, ecosystems present the foundation for everything we do, including all economic processes. This is why it is absolutely fundamental to adopt an ecosystem approach following the CBD and develop new hubs of scientific knowledge in Indonesia, focusing on ecosystem services, thereby increasing resilience and addressing the issues of transparency and traceability that have been abundantly flagged in the literature on biodiversity offsets presented in the following chapter. Without knowing which fragments of ecosystems are most valuable, the introduction of ill-informed biodiversity offsets could harm Indonesia, as opposed to improving the situation. This is why we highly recommend conducting a comprehensive ecosystem services mapping and applying recommended procedures in setting up a system of biodiversity offsets in Indonesia. This way we could follow scientific evidence (Jacob et al., 2016) [21] and create a bespoke system of biodiversity offsets for Indonesia, one of the most biologically diverse countries on Planet Earth. Most of the biodiversity offset schemes created without this step around the world have failed, which the next chapter amply illustrates.

The obvious prerequisite to any successful system of biodiversity regeneration is the information base. Following a most representative classification of ecosystem services, e.g., the Common International Classification of Ecosystem Services, a fully-fledged mapping of ecosystem services in the context of Indonesia needs to be performed. Following the ground-breaking research of (Shmelev et al., 2023) [7], and the multidimensional mappings illustrating the availability of detailed data for France on a 1 km × 1 km grid, this seems to be feasible and of high potential value for this analysis for Indonesia.

It is important to underline that for most of the ecosystem services available in the classification, approximately 30 layers have to be covered for a comprehensive reading of the true value of ecosystems in Indonesia. Some of the most recent publications covering ecosystem services mapping in Indonesia we could mention include (Damastuti & De Groot, 2019; Nugroho et al., 2022; Mathys et al., 2023; Triana & Wahyudi, 2020; Fauzi et al., 2023; Perdinan et al., 2024) [73,74,75,76,77,78].

In the context of Indonesia, further research will have to be carried out to map ecosystem services fully. The first step in this research is going to be the identification of all the sources of spatial information for Indonesia, representing various ecosystem services. Next, the integration of various layers could be conducted with varying weights, followed by the calculation of hotspots.

The most important systems science conclusion we could make for the benefit of this paper echoes (Norgaard, 2010) [79] in the sense that urgent research is required into dynamic interactions between ecosystem services, the often hidden feedback loops and the trade-offs present. This is hugely important in the context of Indonesia, one of the most biologically countries on the planet as we simply often do not know the full extent of the implications of losing a particular fragment of the natural ecosystem say, e.g., on the island of Borneo or the true implications of creating a new road. The new IPBES report (Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services, 2022) [80] fully confirmed a policy shift away from purely monetary towards a more multidimensional assessment of ecosystems.

The failure to meet the Aichi Biodiversity Targets underscores the inadequacy of current global efforts to halt biodiversity loss. Biodiversity offset schemes, which often promise “No Net Loss”, have failed to deliver on this goal for at least 146 years, according to some studies. This failure is rooted in the flawed assumption that ecosystems are fungible commodities. Current approaches are unscientific, lack transparency, and frequently undervalue ecosystems’ complexity and intrinsic contributions to humanity.

Indonesia, one of the world’s most biodiverse nations, has experienced catastrophic deforestation, reducing forest cover from 65% in 1990 to below 50% today. This decline is driven by palm oil production, coal mining, population growth, and other key factors identified and validated through our quantitative analysis. These activities not only devastate ecosystems but also exacerbate economic, social, and ecological vulnerabilities.

Biodiversity offsets, as they are currently implemented, fail to address the complexity of ecosystems. They rely on arbitrary ratios, lack ecosystem mapping, fail to ensure “No Net Loss,” and offer inadequate compensation for lost biodiversity. Replacement ecosystems are often poor substitutes for what has been destroyed, with significant social and cultural costs. Moreover, biodiversity offsets in their current form risk commodifying nature, turning vital ecosystems into mere transactions within neoliberal frameworks.

To address these challenges, we propose the following measures:

• Acknowledge Indonesia’s Global Significance for Biodiversity: Indonesia is one of the most important biodiversity hubs globally, and preserving its ecosystems must be a national and international priority.
• Address the Drivers of Deforestation: Palm oil production, coal mining, and population pressures have been major contributors to forest loss. Effective policies to curb these pressures are urgently needed to reverse the trend of deforestation.
• Implement Comprehensive Ecosystem Services Mapping: Conduct nationwide mapping of ecosystem services at a resolution of 1 × 1 km or finer to identify critical biodiversity hotspots.
• Protect High-Value Areas: Exclude the top 25% of areas with the highest multidimensional biodiversity value from all economic and infrastructure development activities.
• Redefine the Economic Framework: Adopt a framework of ecological economics, placing the economic system firmly within environmental boundaries. Recognize ecosystems as holistic systems with intrinsic and instrumental value.
• Foster Transparency and Accountability: Establish robust monitoring systems in collaboration with Indonesia’s Statistical Office, Geospatial Information Agency, and Space Agency to ensure transparency and long-term tracking of biodiversity offset outcomes.
• Develop Centers of Green Economic Growth: Focus on low-resource, high-value industries such as software development, education, health, eco-tourism, and financial services. Actively pursue ecosystem restoration and regeneration alongside economic development.
• Strengthen Research and Capacity Building: Expand research centers focused on satellite imagery and ecosystem modeling, fostering local expertise in sustainable development and ecological management.
• Position Indonesia as a Global Conservation Leader: By adopting an innovative, data-driven approach, Indonesia can lead mega-diverse nations in developing conservation strategies that align with global biodiversity goals.

6. Conclusions

The research carried out in this paper has reviewed a substantial amount of evidence-based scientific literature on the subject of biodiversity offsets. Various studies have been analyzed from the point of view of critical factors that preclude existing instruments from operating effectively. The following five clusters of factors have been identified: institutions, data, ecology, economy and society. Based on meticulously collected data and the PRISMA-inspired literature review, we have built a multi-factor econometric model of deforestation in Indonesia, covering the period between 2000 and 2020. The following factors have been revealed as statistically significant drivers of deforestation in Indonesia: population growth, palm oil production, coal extraction, with GDP having a mild stabilizing effect. We conclude that the deforestation situation in Indonesia, one of the most biodiverse countries on the planet, is rather serious and that new and innovative science-based tools need to be created instead of the unworkable biodiversity offsets to stop and reverse the loss of biodiversity. The paper outlined specific recommendations that all biodiversity-rich countries could follow and presented food for thought and a call to action in this important area. This paper presents but a first step in a significant body of research that still needs to be done.