A criminal justice response to address the illegal trade of wildlife in Indonesia

The global illegal wildlife trade (IWT) is a multibillion dollar annual trade that threatens numerous species. Understanding ways to improve the law enforcement response is an essential component in addressing this trade. Yet, quantifying the impacts of such conservation measures is often hindered by a lack of long‐term and reliable datasets. Here, we evaluate a 15‐year multistakeholder collaboration that aimed to detect, report, and robustly respond to IWT across the vast Indonesian archipelago. Our results demonstrate the performance of site‐based monitoring networks in reliably reporting a widespread IWT of hundreds of nationally protected species. It revealed highly responsive government law enforcement agencies, high prosecution and conviction rates, and increasing penal sanctions over time, which significantly differed by province, year of arrest, and the number of unique protected species seized in a case. From these results, we formulate management recommendations for key agencies working in the criminal justice system.

policy reforms (Bennett, 2011;Duangchantrasiri et al., 2016). These efforts have had mixed success, leading to recent calls for an interdisciplinary approach toward tackling wildlife crime that learns from criminology research (Boratto & Gibbs 2021).
Criminology studies, albeit from the United States and Europe, show that when applying a criminal justice response to a range of crime types, increasing the certainty of apprehension and punishment rather than the severity of punishment has a stronger deterrent effect on reducing criminal behavior (Nagin & Pogarsky, 2001;Petrich et al., 2021). Longer prison sentences may even be counterproductive by increasing recidivism in offenders who become institutionalized, lose social ties, and lose legitimate employment opportunities, all of which are important for reformation (Gendreau et al., 1996). To understand wildlife crime, studies have tended to focus on determining the drivers, types, and volumes of species trafficked, the crossover between the legal and illegal trade, patterns of species declines, trade routes, including online trade, and predicting species at risk of trafficking (Roberts & Hernandez-Castro, 2017;Scheffers, Oliveira, Lamb, & Edwards, 2019;Latinne et al., 2020;Nijman et al., 2022). Whilst useful in bringing attention to this critical issue and providing insights into changing patterns of trade, quantitative studies that rigorously assess the effectiveness of counter-wildlife trafficking interventions are urgently needed to understand varying deterrence-based responses, as the following case studies show.
To tackle manta hunting and gill trafficking in eastern Indonesia, government authorities identified the geographical and social connectivity of actors in an IWT network operating across multiple islands (Booth et al., 2021). These findings were used to inform the design and implementation of an integrated strategy that targeted key nodes along the IWT chain. Responses ranged from conducting marine patrols to reduce supply at the site-level to confiscating manta gills and prosecuting the main traders in a notorious urban trafficking hub. These coordinated efforts significantly reduced manta hunting and trade. A second case study, on large mammal hunting preferences in the Democratic Republic of Congo, showed that a national wildlife protection law designed to prevent the exploitation of vulnerable species had not achieved this aim, stressing the need for site-based management interventions, such as patrolling, to better protect these species (Rowcliffe et al., 2004).
In our study, we aim to add to this limited body of knowledge by focusing on the IWT in Indonesia, an archipelago nation of over 17,000 biodiverse and accessible islands, and a major source country for this illicit trade (UNODC, 2020). Indonesia has a clear policy framework that regulates wildlife trade through an overarching bio-diversity conservation law (UU5/1990, MoEF, 1990 with derivative regulations on the legal trade (PP8/1999 andKepMenhut447/2003) and wildlife protection (PP7/1999). Still, despite this, there are several main barriers to combatting wildlife trafficking, which are commonly shared by other IWT-affected countries. They relate to limited capacity of law enforcement agencies in specialist areas, such as mapping wildlife trafficking networks to facilitate the detection, arrest and/or prosecution of major poachers and traffickers (UNODC, 2020). To tackle IWT in Indonesia, substantial effort has been invested in establishing well-trained government-led multistakeholder partnerships operating in trafficking hotspots across the country, starting in 2003. This provides a rare opportunity to assess how this intervention, which is commonly used by many other IWT source countries, performed in practice.
We use a 15-year dataset spanning the Indonesian archipelago to assess the: performance of a multistakeholder approach to monitor, report and mitigate IWT; and, outcome of court cases based on information provided by local partners and the factors that explain differences in penal sanctions. Our intervention strategy specifically targeted high priority biodiversity-rich sites and certain at-risk "protected species," as defined under Ministry of Environment and Forestry (MoEF)'s PP7/1999, potential biases that are controlled for in our analyses. Our study did not focus on the unsustainable legal trade or illegal trade in nonprotected species, which are harvested and traded outside of the legal quota system, and which would therefore require a different intervention strategy.

Data collection and compilation
Our dataset is derived from reports made by three multistakeholder networks. This involved creating systems of communication and coordination to channel IWT reports from forest-edge communities (in target landscapes), coastal communities (in target seascapes), and local organizations in urban areas (with trade routes connected to these land/seascapes) directly to a government partner or via an NGO intermediary. Networks were formed to cover central Indonesia (established in 2003), western Indonesia (2003), and eastern Indonesia (2015) and strengthened over time through the addition of NGO field staff to handle an increasing number of reports. Field reports were verified by an NGO employee telephoning the source of information and/or visiting the site to verify a report. Information deemed reliable was submitted to a government contact from the MoEF and/or police, who would decide whether to follow up on a report. Where this elicited a law F I G U R E 1 (a) Number of illegal wildlife trade cases from different geographical areas across the Indonesian archipelago that were reported by the three field teams to government partners for follow up; (b) number of cases dealing with illegally traded priority species or species groups; (c) number of cases differentiated by crime type; and (d) status or outcome of illegal wildlife trade arrests in all target areas combined. The dashed line shows annual increases in the number of field staff enforcement operation, each case would be tracked to determine its outcome, whether proceeding to court and being sentenced (prison time, fine, both or neither), not proceeding to court (and receiving an administrative sanction and having the contraband confiscated) or neither. The final dataset used in the analysis consisted of cases monitored from January 2004 to December 2018, with the latest update on court case outcomes being September 2019. The dataset contains the following case-specific information: date of law enforcement operation; trafficked species name/s, body part/s and quantity; crime type; province where the IWT incident occurred and district where suspect/s were apprehended; verdict for cases pro-ceeding to court (fine, prison time or neither) or outcome for those not proceeding to court (administrative sanction, confiscation letter, extradition, warning letter or case stopped); and, different stakeholders/agencies that first reported the case or supported the development of the case before it proceeded to court. We collected additional IWT case data from the Government of Indonesia's court "Case Tracking Information System" for our subsequent quasiexperimental analysis. Case data were available from 2016 onward, and we collected data from 2016 to 2018 to match our cases from those years.
A case is defined as a single IWT event that is reported and responded to through the project partnership. Most cases consisted of more than one species (mixed species groups) that had varying protection status and quantity. To capture this diverse information, we: recorded the number of unique protected species seized per case (Table S1); and, the monetary value of the total volume of protected species (i.e., species listed on MoEF PP7/1999) traded in a case. We estimated the monetary value of the evidence seized per case using data gathered on the price of different body parts for individual species at the trader level. We multiplied the price per body unit by the number of body parts for a particular species (using specific body part seized). For example, two tiger skins multiplied by USD3664/skin equaled USD7328 (Supporting Information A).
For each case, we calculated the number of different crime types and suspects arrested in each location. Crime types were categorized according to the prevailing MoEF law (UU5/1990), which includes illegal trading, smuggling, and others (illegal hunting, illegal fishing, illegal keeping, illegal ownership, and transportation).

Data analysis
Our data analysis consisted of several sections. To understand the temporal trend (years) in the number of incidents (cases) reported and its association with the changing number of staff and of new field teams (2004-2018), we used a linear regression analysis, after verifying normality of residual distribution. We also investigated (Chi-squared test) whether there was a temporal association between case handling (case stopped, administrative sanction, fine/prison sentence, or outcome not yet known). Reporting, investigations and law enforcement operations in our study targeted certain types of protected species (e.g., tiger) or protected species groups (e.g., parrots and turtles). The confiscations were not, therefore, from investigations that specifically targeted all species in trade, which would have included other heavily traded species groups, such as protected songbird species. This meant that certain species groups were disproportionally represented in our dataset. We do not, therefore, present our results as being indicative of the general pattern of species being illegally traded in Indonesia. Instead, we are interested in using our data to investigate which combination of factors best explain court case sentencing (as the dependent variable). Generalized least square models were constructed for the following variables, after controlling for collinearity, for each case: province where case was tried; crime type; year of arrest; number of unique protected species seized in a case (to minimise the species group bias in the dataset); and, monetary value of protected species seized (log-transformed) (Table S1, Supporting Information B).
Since the dependent variable consisted of two different units-prison time (in months) and fine (in Indonesian Rupiah-IDR, adjusted for annual inflation)-we constructed a single dependent variable (combining prison time and fine) using the following approach. First, we conducted a principal component analysis ( Figure S1) and inspected the loadings of the two variables (months and IDR). The equal loading values of the variables showed that both variables did not need further weighting. So, for each case, the two variables were log transformed, standardized using a scaling function in R and centralized, after which they were added together. Generalized least square models, controlling for heterogeneity of the residual variances (Figures S2-S3) were run using a restricted maximum likelihood approach and model selection based on a second order Akaike information criterion adjusted for small sample sizes (AICc) ( Table S2).
We used a propensity score matching (PSM) technique to match project-assisted cases (treatment) versus nonproject cases (control), taken from the Indonesian government's "Case Tracking Information System." Matching was done based on treatment and control cases having the F I G U R E 3 Mean coefficient estimates (95% confidence intervals) of explanatory variables for the sentence index of project assisted cases from 2004 to 2018. Significant variables are denoted with a blue dot and the adjusted %R 2 is 43.71%. Significance level is at p value < 0.05 following shared case features: province where case was tried; crime type; year of arrest; number of unique protected species seized in a case; and, monetary value of protected species seized (log-transformed). The technique generated paired cases and omitted treatment cases that had no accurate match. We tested the most efficient PSM method by considering improvements in the mean difference between the treatment and control groups and the number of paired cases generated ( Figure S4). Using the generated paired cases, we tested for differences in sentencing index scores between the treatment and control groups (Wilcoxon rank-sum-test). We investigated tempo-ral patterns of case reporting by nonstate and state agencies and the involvement of different government agencies (Chi-squared test).

Patterns of IWT
Over the 15-year study, information was compiled from 390 cases that were reported through the project partnership and involved 228 unique species or species groups, of F I G U R E 4 Propensity-score matching estimates on the effect of the multistakeholder partnership in eliciting higher penal sentences (average treatment effect on the treated) from 2016 to 2018 which 64.9% were on the national protected species list and therefore being illegally traded. There were 170 species on the IUCN Red List, with 85 threatened species (CR = 10.0%, EN = 15.3% and VU = 24.7% of the total). Eleven species or species groups represented the majority (66.4%) of the trafficking cases, mainly for tiger (19.7%), parrots (9.2%), turtles (8.7%), Sunda pangolin (5.9%), and manta ray (5.9%; Figure 2b). The number of cases was significantly and positively correlated with an increase in field effort from one to three teams, and three to 33 field staff (r s = 0.55 and 0.80, respectively, p < 0.05; Figure 1a). From 2004 to 2018, the number of cases compiled increased from 29 to 74 and the cumulative number of unique species/species groups confiscated increased from 28 to 228 (Figure 1b).
The majority (54.4%) of wildlife crime types were trade, which disproportionately increased over time as compared to the prevalence of other crime types (Figure 1c). Online trade was first identified in 2011 (one case) and by 2018 involved 23 cases (31.1% of the trade cases). The IWT cases changed from being issued with an administrative sanction to those arrested (75.9% in 2004), to proceeding to court and receiving a sentence (67.6% in 2018; χ 2 = 93.62, p < 0.01; Figure 1d).

Patterns of fines and prison sentences
The 390 reported cases led to the arrest of 474 suspects, of which 376 cases involving 456 suspects had a known outcome. Most (59.8%) cases proceeded to court and, of these, most (80.0%) received a fine and prison sentence (Figure 2). The total prison time issued was 2723 months (mean = 12 months ± 10, SD), with USD411,989 levied in fines (mean = USD1831 ± 2605). The estimated monetary value of evidence seized in these cases was USD12,280, 175.
Sentencing outcome was significantly influenced by province, year of arrest and total number of unique protected species seized in a case (Figure 3). Cases tried in East Java, Maluku, and West Nusa Tenggara received lower than average sentences. Prosecuted cases were mostly related to illegal trade (70.7%) and smuggling (18.2%) and others (4%). Average sentences increased over time and were also higher for cases involving a greater number of high priority species (Figure 3). Our matching technique paired project-assisted cases with an equal subset of control cases, based on the court case outcome covariates, and found that cases assisted by the multistakeholder partnership elicited higher penalties (W = 5740, p < 0.05; Figure 4).

Role of the multistakeholder partners
From 2004 to 2014, nonstate agencies were more frequently involved in reporting cases, but from 2017 onward stateagencies were (χ 2 = 118.55, p value < 0.001; Figures 5a), and with a greater average number of partnering agencies per case (from 1.0 ± 0.0, ± SD, to 2.6 ± 1.5). The MoEF was the most active partner throughout the study, while border security and the National Police became more actively involved after 2014 (χ 2 = 189.73, p value < 0.001; Figures 5b).

DISCUSSION
As the conservation community grapples with how to tackle unprecedented levels of wildlife trafficking (Challender & Macmillan, 2014;Phelps et al., 2014), our evidence-based study demonstrates the performance of a government-led multistakeholder enforcement approach in a major IWT source country. Our partnership was able to detect wildlife trafficking cases across the Indonesian archipelago and gather reliable information for government law enforcement agencies to act upon. Establishing teams and partnerships to monitor IWT in new geographies resulted in additional wildlife seizures and arrests by government, thereby illustrating the high replicability and impact of this approach. Another essential component of the strategy was having highly motivated and capable government law enforcement agency partners, who responded F I G U R E 5 The dynamics of state and nonstate organizations involvement throughout the years to 63% of the 390 reports submitted, of which most were prosecuted. We discuss our study limitations, key findings, knowledge gaps and make recommendations to enhance the strategy. Our dataset contained temporal biases that were associated with the number of cases handled, type of species confiscated and crime type prevalence. The 376 cases analyzed over 15 years also represents a detailed analysis of aspects of IWT in Indonesia. For example, there were an additional 430 IWT cases handled by government agencies for 2016-2018, whereas our dataset contained 172 cases for this period, and online IWT monitoring has revealed thousands of advertisements selling protected species in Indonesia, which were not captured in our study (e.g. Fink et al., 2021). We stress the importance of future studies analyzing these larger data sets to further understand IWT patterns and interventions. Regarding crime types, the teams initially gathered information on illegal wildlife ownership, but these cases received weaker (mainly administrative) sanctions because they were perceived as a less serious offence by the authorities. From 2008 onward, the strategy shifted to addressing trafficking cases. This expanded the multiagency partnership and increased cooperation with nonstage agencies and resulted in substantial wildlife seizures. It marked a turning point and from 2014 onward two salient trends emerged: most cases handled proceeded to court; and, the penal sanctions increased. One explanation for this finding is that a greater number of government agencies, with their own exper-tise and resources, partnered with MoEF to jointly tackle a complex and widespread IWT in Indonesia.
Efforts to increase the number of cases proceeding to court and prosecuted was bolstered through counterwildlife trafficking government trainings and, in 2017, the Attorney General instructed all senior prosecutors to prioritize IWT case handling. This led to 212 state attorneys across Indonesia being trained. To formalize this approach, accredited IWT training modules were developed for the National Police (in 2017), Supreme Court (in 2017), and Attorney General's Office (in 2019). This growing partnership, trainings, technical assistance in court case preparation, and the greater attention to prosecuting IWT cases, explains why project-assisted cases elicited higher sentences.
We found disparities in penal sanctions issued for project-reported cases. For example, a trafficker prosecuted in either of the provinces of East Java, Maluku, and West Nusa Tenggara would receive, on average, a penalty almost eight times lower than the average sentences in the other Indonesian provinces after controlling for year of arrest and other factors. Perhaps a reflection of the lower levels of investment in partnerships and government agency capacity in these provinces than elsewhere. Alternatively, sentencing outcome may have been related to local contextual features, such as court caseload pressure, sentencing norms, or local prison capacity, which were not measured in our study (Ulmer & Johnson, 2004). Higher sentences were issued for cases involving multiple high priority species, such as tiger, Sumatran orangutan, elephant, and yellow-crested cockatoo, suggesting that the greater conservation attention afforded to these species had an influence.
In our study, what remains unclear is whether higher rates of prosecutions and sentences had a deterrent effect and, if not, the threshold required for achieving this. We recognize that to rigorously evaluate strategy effectiveness would require demonstrating that the intervention slowed or reserved species population decline due to IWT or that it created a deterrence effect. This would be hugely challenging given the variety and widespread geography of species being traded. However, our study provides encouragement that the preconditions of deterrence theory (namely certainty and severity of punishment) are being met.

Recommendations
Recognizing the importance of certainty when deterring offenders, we recommend measures to better identify priority offenders and activities to increase this certainty. These would include greater support for interagency approaches to jointly tackle wildlife crime, such as joint trainings and enhanced criminal mapping to improve the detection and targeting of key offenders, especially traders and smugglers, as part of a preventative strategy. Our approach demonstrates higher success in increasing punishment certainty. Publicizing successful convictions through greater media exposure and community meetings, would increase awareness of the certainty of punishment, but also of the penalties themselves, thereby increasing deterrence as part of a persuasive approach. In Indonesia, the maximum sentence of 5 years in prison and/or a USD7,500 fine seems disproportionately low compared to several high-end seizures valued between USD156,976 and USD366,195 (involving helmeted hornbill casque, pangolin scales, and Asian elephant tusks). However, it is unlikely that increased prison sentences would deter such cases in the future and there is strong evidence, from studies on other crime types, against increasing prison sentences to reduce criminal behavior (Gendreau et al., 1996). While this association remains untested for IWT, it likely holds true also (Wilson & Boratto, 2020). We therefore recommend a focused deterrence approach that targets high-risk individuals with interventions that disincentivize them, such as increased surveillance (Apel & Nagin, 2011).
Our study highlights key areas for future research that should enable a better understanding of the response required to target the major drivers of IWT. This includes developing analytical techniques to estimate criminal network size for different species group, what proportion of this network our study represents and how far our prose-cutions have gone toward dismantling it (Indraswari et al., 2020). Research is needed into motivation to participate in IWT, awareness of punishment risks (Paudel, Potter, & Phelps, 2020), whether longer or shorter penal sanctions reduce criminal behavior, whether prison sentences are counterproductive because they inadvertently radicalize offenders (Nagin, 2013), and to better understand why prosecutions varied amongst provinces. These are complex questions to address and require further exploration through a multidisciplinary evidence-based approach.

A C K N O W L E D G M E N T S
This study was conducted under the MoU between the Indonesian MoEF and WCS, for which WCS thanks all MoEF staff involved for their support. This work was partially funded by the Liz Claiborne and Art Ortenberg Foundation, US Fish and Wildlife Service, Foundation Segrè, AZA Tiger Conservation Campaign, DEFRA, UNDP/GEF, US Department of State's Bureau of International Narcotics and Law Enforcement Affairs, USAID, Morgan Family Foundation, Disney Conservation Fund, the Critical Ecosystem Partnership Fund and Vulcan/Paul G. Allen Family Foundation. We thank Irma Hermawati for data collection, Eleni Matechou, Aiden Keane, and Tancredi Caruso for statistical advice, and Scott Roberton, Jonathan Hunter, and three anonymous reviewers for commenting on the manuscript.