Seller’s reputation and capacity on the illicit drug markets : 11-month study on the Finnish version of the Silk Road

1 This is a pos p in e sion o he a icle: Nu mi, J., Kaskela, T., Pe älä, J., & Oksanen, A. Selle ’s Repu a ion and Capaci y on he Illici D ug Ma ke s: 11-Mon h S udy on he Finnish Ve sion o he Silk Road. D ug and Alcohol Dependence, 2017, ol. 178, pp. 201-207. DOI: 10.1016/j.d ugalcdep.2017.05.018. h p://www.d ugandalcoholdependence.com/a icle/S0376-8716(17)30291-0/pd Selle ’s Repu a ion and Capaci y on he Illici D ug Ma ke s: 11-Mon h S udy in he Finnish Ve sion o he Silk Road Abs ac Aims: This 11-mon h s udy analysed illici d ug sales on he anonymous To ne wo k, wi h a ocus on in es iga ing whe he a selle ’s epu a ion and capaci y inc eased daily d ug sales. Design and se ing: The da a was ga he ed om Silkki ie, he Finnish e sion o he Silk Road, by web c awling he si e on a daily basis om No embe 2014 o Sep embe 2015). The da a includes in o ma ion on selle s (n=260) and p oduc s (n=3823). Measu emen s: The measu emen s include he selle s’ epu a ion, he sale amoun s (in eu os), he numbe o a ailable p oduc s and he ypes o d ugs sold. The selle s’ capaci y was measu ed using hei ull sales po en ial (in eu os). Fixed-e ec s eg ession models we e used o es ima e he e ec s o selle s’ epu a ion and capaci y; hese models we e adjus ed o he ypes o d ugs sold. Findings: O e all, illici d ug sales o alled o e 2 million eu os du ing he s udy, bu many p oduc s we e no sold a all, and selle s we e ac i e o only a sho ime on a e age (mean=62.8 days). Among he p oduc s sold, s imulan s we e mos widely pu chased, ollowed by cannabis, MDMA, and psychedelics. A selle ’s epu a ion and capaci y we e bo h associa ed wi h d ug sales. Conclusion: The To ne wo k has enabled a ans o ma ion in d ug sales. Due o he ne wo k’s anonymi y, he selle ’s epu a ion and capaci y bo h ha e an impac on sales. Key wo ds: illici d ugs, c yp oma ke , In e ne , To ne wo k, anonymi y, longi udinal, web c awling 1. In oduc ion Illici online d ug sales i s a ac ed majo media a en ion wi h he ise and all o he Silk Road in 2011–2013 (Ba a e al., 2014; Ma in, 2014a; Ma in, 2014b). The Fede al Bu eau o In es iga ion shu down he Silk Road in Oc obe 2013 and i s successo , he Silk Road 2, in No embe 2014 (Dolli e , 2015). The Silk Road 3 was launched almos immedia ely a e bu was o e un by o he si es, such as Ago a and E olu ion, which adop ed some ea u es o he o iginal Silk Road (Dolli e e al., 2016; Van Buski k e al., 2016). The common ea u e o all o hese websi es is ha hey p o ec he anonymi y o hei use s; hey a e hence e e ed o as “c yp oma ke s” (Ald idge and Déca y-Hé u, 2016; Banc o and Reid, 2017; Deman e al., 2016; Ma in, 2014b; Van Buski k e al., 2016). This a icle ocuses pa icula ly on he Finnish online se ice called Silkki ie. Silkki ie ope a es using he To ne wo k o hide i s eal loca ion, jus as Ago a and a ious Silk Road e sions did. The To (which s ands o The Onion Rou e ) ne wo k and so wa e a e designed o de end 2 human igh s (Dingledine, 2004). This ne wo k can be used o censo ship ci cum en ion, online anonymi y and high-le el online p i acy. I is possible o hos a websi e inside he To ne wo k (a so- called onion se ice) and o hide he physical loca ion o he si e’s se e . To p o ides anonymi y by ou ing he use ’s a ic h ough h ee sepa a e elay se e s so ha i is di icul o e eal he use ’s physical loca ion o IP add ess. This means ha To p o ec s use s by enc yp ion o ensu e p i acy, au hen ica ion be ween clien s and elays and signa u es o ensu e ha all clien s know he same se o elays. To is conside ed e y secu e and esilien agains su eillance: The U.S. Na ional Secu i y Agency called To “ he king o high-secu e, low-la ency In e ne anonymi y” (The Gua dian, 2013) in i s op-sec e documen a ion, in which i discusses i s u ile a emp s o spy on To use s. Despi e his, di e en police ope a ions ha e a ge ed si es and ha e a leas been pa ially success ul in educing he ac i i y in c yp oma ke s (Déca y-Hé u and Giommoni, 2017). Also, he iden i y o he o iginal Silk Road de elope , Ross William Ulb ich, was disco e ed a e he FBI ollowed his beha iou pa e n and ga he ed cumula i e c i ical in o ma ion o long enough (FBI, 2014). The anonymi y o online d ug ma ke s is gua an eed by digi al cu encies and paymen sys ems ha signi ican ly inc ease he di icul y o acking down buye s and selle s. Bi coin is one digi al cu ency and a paymen sys em in one; i was in en ed by an unknown en i y who calls himsel Sa oshi Nakamo o (Nakamo o, 2008; Bohannon, 2016; K is ou ek, 2015). Bi coin is a pee - o-pee applica ion ha allows digi al money ans e s be ween use s. I s mone a y sys em is decen alised and designed o wo k wi hou cen al banks, go e nmen s and egula ions. Al hough Bi coins a e used in mul iple legal ways, hese anonymi y p ope ies also make Bi coin a ac i e o c iminals. The wo k on illici c yp oma ke s has expanded apidly since 2015. Many o he s udies ha e been based on he Silk Road (e.g., Al idge and Déca y-Hé u, 2016; Ba a and Maddox, 2016) and hei successo s, such as he Silk Road 2.0, Ago a, Alphabay and Valhalla (e.g., Van Buski k e al., 2016; Déca y-Hé u and Giommoni, 2017; Deman e al., 2016; Dolli e and Kuhns, 2016; Van Hou and Hea ne, 2017). S udies ha e employed mo e adi ional esea ch me hods such as su eys and in e iews (Ba a e al., 2014; Van Hou & Bingham 2013a, b) and newe app oaches employing web c awle s (Al idge and Déca y- Hé u, 2016; Deman e al., 2016; Dolli e , 2015; Dolli e e al., 2016; Ald idge, 2016; Ch is in, 2013; Ha dy and No gaa d, 2015; Munksgaa d e al., 2016). Some s udies ha e also in es iga ed he quali y o d ugs sold in he c yp oma ke s (Caude ille e al., 2016; Rhumo ba be al al., 2016). The e is also cu en ly a need o unde s and ha c yp oma ke s a e also used e y locally, as in ou Finnish case. So a coun y cases ha e in ol ed, o example, Canada (B oséus e al., 2016) and Swi ze land (Rhumo ba be e al., 2016), and some s udies ha e in ol ed c oss-na ional compa ison (Van Buski k e al., 2016). S udies on he Silk Road indica e ha use s we e ypically males in hei 20s and p io i ised he Silk Road o e s ee ma ke s o quali y easons and o pe sonal sa e y (Ba a e al., 2014; Van Hou and Bingham, 2013b; Ba a e al., 2016a, Ba a e al., 2016b). C yp oma ke use s ha e been cha ac e ised as a “ echnological d ug subcul u e” (Van Hou and Bingham, 2013b), which also ep esen s a o m o online ac i ism unde lining indi idual eedoms based on libe a ian e hos (Maddox e al., 2016; Munksgaa d and Deman , 2016). Use s’ igh o choose is combined wi h he exp essed joy o ha ing he oppo uni y o choose hei p e e ed d ugs like “kids in a candy s o e” (Ba a e al., 2016b). As c yp oma ke s a e o ganised o gua an ee he ull anonymi y o he use s, hey also b ing challenges o he use s. Social psychologis s ha e unde lined ha anonymous beha iou can be highly egula ed on he In e ne (Keipi and Oksanen, 2014; Spea s e al., 2002). In c yp oma ke s, anonymi y is en o ced by ins abili y, which may e en en o ce he anonymi y e ec s o online beha iou . Ins abili y is one o he 3 cen al ea u es o online d ug ma ke s, as mos i ems a e sold quickly and he majo i y o selle s disappea wi hin a ew mon hs (Ch is in, 2013). Hence, he e is a need o building “ epu a ion sys ems” (Resnick, 2000; House and Woode s, 2006) o “ us sys ems” (Lus haus, 2012) on c yp oma ke s. Public use eedback is one o he common pa s o such epu a ion o us sys ems. A p e ious s udy by Ha dy and No gaa d (2015) showed ha selle epu a ion had a posi i e impac on cannabis sales. They a gue ha use s’ eedback is a cen al ea u e in cannabis sales on he Silk Road because selle s and buye s do no know each o he . Posi i e epu a ion gi es he imp ession ha he selle is us wo hy, which leads o inc eased sales (Ha dy and No gaa d, 2015). I has also been ound ha endo s who had poo use a ings we e mo e likely o ake he isk o shipping d ugs in e na ionally (Déca y-Hé u e al., 2016). Besides use eedback, i is o en impo an o communica e well wi h he cus ome and c ea e an imp ession o us wo hiness. Acco ding o Déca y-Hé u and Leppänen, success ul online c iminals ha e communica i e skills ha lead o highe ewa ds (Déca y-Hé u and Leppänen, 2013). In he absence o isual cues ha a e ypically a ailable in an o line con ex , online buye s mus ely on wha in o ma ion is a ailable, which also sepa a es online c yp oma ke s om he adi ional d ug ade. We a gue he e, in addi ion o he selle ’s epu a ion, ha selle ’s apaci y is ano he po en ial ac o o success ul sales. Selle ’s capaci y e e s o selle s’ p omises ega ding he a ie y and quan i y o high- quali y p oduc s a ailable. This is pa o communica ion wi h he cus ome s. Selle ’s capaci y o p omise o a lo o d ugs in s ock will c ea e posi i e expec a ions ha will lead o highe sales. I is common o cybe c iminals o cons uc an online iden i y and c ea e us wo hiness o a ac c iminal pa ne s (Lus haus, 2012). Selle ’s capaci y is pa o his communica ion, bu i has no been p e iously s udied wi h ega d o c yp oma ke s. Howe e , economic s udies on ad ance selling ha e shown ha selle capaci y is posi i ely associa ed wi h consume beha iou (Yu e al., 2015). In his s udy, ou aim was o esea ch bo h he selle ’s epu a ion and capaci y in Silkki ie. We will i s p o ide desc ip i e esul s on he gene al d ug sales on Silkki ie. The main ocus o he a icle lies, howe e , on he epu a ion and capaci y ha ep esen di e en ace s o he selle ha a e a ailable o buye s. Anonymi y di ec s use s’ a en ion o he a ailable ea u es and ypically also en o ces g oup beha iou (Keipi and Oksanen, 2014; Spea s e al., 2002). Hence, in c yp oma ke , us sys em is based on bo h di ec eedback and success ul communica ion (Ha dy and No gaa d, 2015; Resnick, 2000; House and Woode s, 2006; Déca y-Hé u and Leppänen, 2013; Lus haus, 2012). Ou hypo heses we e g ounded on hese p emises, and we expec ed ha a posi i e selle epu a ion would be posi i ely associa ed wi h ha en i y’s daily sales. In addi ion, we expec ed ha an inc ease in a selle ’s capaci y would inc ease ha en i y’s sales. 2. Me hods 2.1. Silkki ie as a ma ke place Silkki ie was opened in he To ne wo k on 6 Janua y 2014, and i was o iginally in ended o be used speci ically o illici d ug sales in Finland. Because bo h he selle and he buye a e loca ed wi hin Finland, e en de ec ing he d ugs is di icul because he e is no cus oms p ocess o sc eening domes ic shipmen s. Fo hese easons, i ually all o he Finnish selle s men ion ha he shipmen s a e pos ed ia domes ic mail. As Silkki ie is he main online ma ke place o Finnish d ug buye s, i is a use ul measu e o he o e all Finnish online d ug ma ke . In 2015, Silkki ie published he English ansla ion o he si e o a ac use space ou side Finland. The o iginal Finnish e sion (h p://silkki iehdg5mug.onion/) now 4 has an English ansla ion called Valhalla (h p:// alhallaxmn3 ydu.onion/). Bi coin cu ency and he Bi coin walle sys em a e used o paymen , and Silkki ie equi es selle s o use PGP enc yp ion (P e y Good P i acy, see Zimme man 1995) and o e hei public keys o buye s o enc yp ed and sa e ex communica ion. Silkki ie shows p ices in eu os and Bi coins using he la es exchange a e. We ha e collec ed ansac ion da a in e ms o eu os. A e Silkki ie use s ha e c ea ed hei use accoun s on he se ice, hey may selec a ole as a selle , which enables hem o o e p oduc s on sale. The selle s may hen se he p ice and desc ibe bo h he p oduc and he e ms o he deli e y. The selle ’s epu a ion is a cen al ea u e on Silkki ie (see Figu e 1). Buye s can gi e eedback by a ing he p oduc and he ansac ion (good: + o bad: –). Buye s see his eedback, so selle s ha e o keep up hei epu a ions. Mo e impo an ly, Silkki ie o e s esc ow se ice by holding he Bi coins used o pu chase and e u ning hem o buye s i he e is a lo o nega i e eedback. This makes i e y impo an o selle s o build good epu a ions. 2.2. Web-c awling p ocess We employed au oma ic web-c awling and web-sc aping echniques o ex ac in o ma ion om he Silkki ie si e e e y day be ween 5 No embe 2014 and 23 Sep embe 2015. This me hod is simila o he one ha was de eloped o s udy he Silk Road’s ansac ions (Ch is in, 2013). F om a wide pe spec i e, we a e using he Open Sou ce In elligence me hodology (Glassman and Kang, 2012). Ou me hod is simple and e y p ecise. Ou web-sc aping so wa e ex ac ed he p oduc in o ma ion om e e y p oduc page on Silkki ie. The collec ed ields we e i le, p ice, numbe o i ems in s ock, selle ’s nickname, selle ’s posi i e and nega i e eedback, selle ’s coun y, buye ’s coun y and he p oduc page’s unique URL. We execu ed he so wa e once a day o e a pe iod o almos 11 mon hs (322 days). Once a day is enough o s udy day- o-day d ug ade. Du ing ou da a collec ion pe iod, 36 days a e missing due o he Silkki ie se e being o line. Hal o hese days ook place in June and July 2015. Du ing his pe iod, Silkki ie use s aised conce n ha he adminis a o o Silkki ie, who calls himsel Kap eeni (“Cap ain” in English), had shu down Silkki ie and un o wi h he emaining Bi coins. The se ice con inued, howe e , a e a sho b eak. 2.3. Measu es Fo his s udy we only used sales ha ook place wi hin Finland by emo ing i ems whe e he selle 's loca ion was ab oad o selle was willing o sell ab oad. The ull da a includes 93 878 obse a ions, and he da a co e s 286 days o ansac ions om 260 selle s o e ing 3823 p oduc s. The da a includes in o ma ion on p oduc s’ p ices, he amoun in s ock and addi ional de ails. D ug sales we e calcula ed using he in o ma ion on a ailable s ock. Fo analy ic pu poses, he da a was o ganised in o panel o ma and included 16 815 daily obse a ions o selle s, including selle ’s daily d ug sales, epu a ion, capaci y and he d ugs ha he selle had sold du ing he speci ic day. Selle ’s daily d ug sales a e he o al alue o all he p oduc s sold in a single day. The median amoun o daily sales was 0, bu he mean daily sales we e 129 eu os/selle . The excess o ze o- alue days is caused by he ac ha many selle s did no make sales on a daily basis. This alue is used as an ou come a iable in he eg ession models. 5 Silkki ie selle s ha e ca ego ised hei p oduc s as s imulan s, cannabis, opia es, empa hogens, psychedelics, opia es, dissocia i es, dep essan s and o he pha maceu icals (see Table 1 o he speci ic in o ma ion). We decided o use his ca ego isa ion, which seems o ep esen expe iences ha buye s y o gain om he d ugs, ins ead o ca ego isa ions based on he chemical o ma ions and legal s a us. A selle ’s epu a ion quan i ies he selle ’s combined posi i e and nega i e eedback ( epu a ion = posi i e commen s – nega i e commen s). The selle s’ epu a ions anged om –19 o 3418. Howe e , ou o 16 815 obse a ions, only 101 (0.6%) we e unde ze o. Based on he 16 815 daily obse a ions o selle s, he median capaci y was 78 eu os. The selle epu a ion igu es o di e en ypes o d ugs a e p o ided in Table 2. A selle ’s capaci y is calcula ed o each day o each selle . I is based on he o al alue o he selle ’s epo ed s ock in eu os in a single day (capaci y= 𝑝𝑟𝑖𝑐𝑒&𝑠𝑡𝑜𝑐𝑘& + &,- ), whe e n is he o al numbe o p oducs ha he selle is o e ing. In o he wo ds, he selle ’s capaci y is calcula ed on he basis o he daily p oduc in o ma ion o each selle . Based on he 16 815 daily obse a ions o selle s, he median capaci y was 1602 eu os. Capaci y igu es o di e en ypes o d ugs a e p o ided in Table 2. 2.4. S a is ical analyses We i s an desc ip i e s a is ics using he ull da a wi h a o al o 93 878 obse a ions. We used S a a 12.0 o un ixed-e ec s Poisson eg ession models o analyse he daily d ug sales (in eu os). These models accoun ed o bo h he selle s and he day when he da a was ga he ed, anging om 1 (5 No embe 2014) o 268 (23 Sep embe 2015). Poisson eg ession was selec ed due o he skewed con inuous ou come a iable. Al hough he e a e se e al al e na i es o ea ing such ou comes (including a linea eg ession wi h log- ans o med ou comes), Poisson eg ession has been a gued o p o ide he bes al e na i e (San os and Ten ey o, 2006). We an he models using Hube -Whi e s anda d e o s (i.e., obus s anda d e o s) due o o e dispe sion (Palme , 2007). Ano he op ion would ha e been using a ixed-e ec s nega i e binomial eg ession, bu he e is isk o inciden al pa ame e bias ha does no occu wi h Poisson eg ession models (Allison, 2012). The models i s epo he unadjus ed e ec s o bo h epu a ion and a ailabili y. We used a squa e- oo ans o ma ion o he selle ’s epu a ion and a na u al loga i hmic ans o ma ion o he selle ’s capaci y o co ec he skewness o hese a iables. The inal models include d ug ca ego ies as dummy a iables (0=no sales in ha ca ego y, 1=a leas one sale in ha ca ego y). Models we e checked o mul icollinea i y; no p oblems we e ound in his espec , and selle s’ epu a ions and capaci ies a e bo h clea and dis inc p edic o s o hei daily sales. The esul s sec ion also epo s he daily sales based on he pe cen iles o he selle ’s epu a ion and capaci y. 3. Resul s The o al alue o he illegal d ugs sold on Silkki ie du ing he 11 mon hs o he s udy was 2 171 387 eu os, and 41 131 i ems we e sold du ing his pe iod (see Table 2). S imulan s we e he mos aluable class o d ug (538 354 eu os), ollowed by cannabis p oduc s (487 111 eu os) and empa hogens (mos ly MDMA; 396 127 eu os). Dep essan s we e he leas sold, wi h only 4975 eu os o sales. Figu e 2 shows 6 he sales alues o di e en subs ances du ing he s udy. We can see ha s imulan s, cannabis and empa hogens combine o make up o e hal o all sales du ing each mon h. Ou o 3823 p oduc s lis ed, only 51% we e sold. Psychedelics had he highes p opo ion o sold p oduc s (64%), and opia es had he lowes (38%). The mean p ice o he sold p oduc s was 84 eu os, and he mean daily sales amoun pe selle was 129 eu os, bu he la ges daily sale by a selle was 85 000 eu os. High igu es in daily sales we e, howe e , e y unusual, as 5000 eu os/day we e exceeded only 10 imes, and by only 7 selle s ou o a o al o 260 selle s. Conside ing all he sales aking place du ing he ollow-up pe iod o he s udy, almos 15% o he selle s did no ha e sales a all, while he highes selling 15% we e esponsible o 82% o he o al sales in Silkki ie. We used ixed-e ec s eg ession models o analyse he associa ions be ween a selle ’s daily sales and his/he epu a ion and capaci y (Table 3). Model 0 epo s he unadjus ed e ec s on daily sales due o a selle ’s epu a ion and capaci y. The models we e un sepa a ely o bo h epu a ion and capaci y, which we e bo h posi i ely associa ed wi h sales. Model 1 includes bo h he selle ’s epu a ion and capaci y in he same model. Model 2 adds he daily sold subs ances as dummy a iables. None o hese dummy a iables we e s a is ically signi ican in he model a he le el o p<0.05; hence, hey we e no included in Table 3. Model 2 included 222 selle s and 16 243 obse a ions in o al, o an a e age o 72.9 daily obse a ions pe selle . The model is s a is ically signi ican ; bo h selle epu a ion and capaci y we e posi i ely associa ed wi h daily sales. Figu e 3 shows he changes in daily sales on Silkki ie based on he pe cen ile alues o he selle ’s epu a ion and capaci y (adjus ed based on a ixed-e ec s eg ession model). The model includes he selle ’s epu a ion and capaci y as p edic o s; i is adjus ed o he daily sales o di e en subs ances. We can see ha he (lowes ) 10 h-pe cen ile g oup in e ms o epu a ion has daily sales o only 62 eu os on a e age and ha he (highes ) 90 h-pe cen ile g oup has sales o 222 eu os on a e age. A simila linea inc ease is seen o selle ’s capaci y. We can also see ha he sales a e especially high o he 80 h and 90 h pe cen ile in e ms o epu a ion. 4. Discussion Ou esul s showed ha he illici d ug sales on Silkki ie we e wo h o e 2 million eu os du ing he 11- mon h s udy. Ou main goal in his s udy was o analyse he impac o he selle ’s epu a ion and capaci y on daily d ug sales. Theo e ically, he a icle was g ounded on p e ious wo k done on us sys ems (Resnick, 2000; House and Woode s, 2006; Déca y-Hé u and Leppänen, 2013; Lus haus, 2012) and p e ious esea ch showing ha he selle ’s epu a ion has a posi i e impac on sales in d ug ma ke s (Ha dy and No gaa d, 2015). We also conside he selle ’s daily capaci y (i.e., he epo ed sales po en ial), which has been shown o ha e an in e ac ion wi h consume beha iou (Yu e al., 2015), bu his e ec has no p e iously been s udied in a c yp oma ke se ing. Ou Silkki ie indings p o ide u he empi ical e idence on he impo ance o us sys ems on c yp oma ke s. Highe epu a ions and capaci ies indica ed highe daily sales. Repu a ion and capaci y ep esen di e en ace s o in o ma ion a ailable o he c yp oma ke consume s, and hey bo h can be conside ed an in eg al pa o he us sys em in he c yp oma ke . The mos success ul selle s ha e bo h a high epu a ion and a high capaci y. Speci ically, he capaci y igu es a e dependen on wha he selle s ha e epo ed as he a ailable s ock and we conside hem as pa o communica ion wi h he buye s. I has been also indica ed in p e ious s udies ha cybe c iminals a e expec ed o a ac companions in c ime h ough success ul communica ion (Déca y-Hé u and Leppänen, 2013). 7 I is possible ha a selle ’s epu a ion and capaci y ha e become e en mo e impo an due o he ins abili y o c yp oma ke s. Wi hin ou da a, we could see ha daily d ug sales a ied conside ably du ing he s udy pe iod, many p oduc s we e no sold a all and ha many selle s emained ac i e o only a sho pe iod o ime. Speci ically, isk-a e se online consume s migh a oid such selle s. A leas in ou da a, his was shown as a concen a ion o sales owa ds hose selle s who had highe epu a ions and capaci y. Resul s on epu a ion and capaci y a e, howe e , also alid o he smalle selle s and sales o a a ie y o d ugs. Silkki ie also p o ides a e y in e es ing na ional case ha d ug ma ke s a e changed by he c yp oma ke s. Al hough s imulan s, cannabis p oduc s and empa hogens (e.g., MDMA) we e among he mos sold p oduc s du ing he s udy, a wide a ie y o subs ances we e sold. This ollows he gene al use pa e n in Finland. The d ug use p oblem in Finland is cha ac e ised by he absence o he oin and by he high p e alence o amphe amine and bup eno phine use (EMCDDA, 2016). Cannabis, amphe amines and MDMA a e he mos widely used illegal subs ances in Finland (Hakka ainen e al., 2015). Compa ed o o he widely popula d ugs, hese subs ances a e mo e associa ed wi h ec ea ional use; opioids and dep essan s a e mo e popula among ma ginalised d ug use s (Pi känen e al., 2016). Based on he was ewa e analyses, he e is an es ima e ha he size o he Finnish d ug ma ke o s imulan s (amphe amine, cocaine and me hamphe amine) was a ound 80 million eu os in 2014 (Kankaanpää e al., 2016). The cannabis ma ke can be es ima ed o be somewhe e be ween 20 million and 100 million based on es ima ions o cannabis use (c. . Hakka ainen e al., 2006), and he e a e ma ke s o bup eno phine and benzodiazepines, used mos equen ly bu no only by disad an aged d ug use s (Tammi e al., 2011). Conside ing his, Silkki ie is a majo ma ke place in Finland. Howe e , mos o he d ugs in Finland a e sold h ough adi ional ou es. I is also possible ha Silkki ie p o ides a o um o new ypes o d ug use s who would no necessa ily buy hei d ugs om he no mal o line ma ke s (Ba a e al., 2016a, 2016b). Acco ding o he Finnish law en o cemen , he la ges endo o Silkki ie, called Douppikauppa, was caugh in Ap il 2016 when a Finnish Cus oms uni in e cep ed a ehicle ha was smuggling d ugs be ween he Ne he lands and Finland (Finnish Cus oms, 2016). The ca con ained se e al kilos o illegal d ugs, including ecs asy able s, LSD abs, amphe amine and me hamphe amine (Finnish Cus oms, 2016). This did no dis up he d ug sales on Silkki ie and Valhalla, howe e . Indeed, he Silkki ie case simply demons a es again how di icul i is o con ol and p e en sales in c yp oma ke s. Al hough law en o cemen could success ully cap u e some o he key ac o s in ol ed, he ma ke s a e quickly e- es ablished (Déca y-Hé u and Giommoni, 2017). Despi e he ac ha we we e able o collec da a o almos 11 mon hs, we would like o men ion some limi a ions o ou s udy. We pe o med da a ex ac ion once a day, bu on some occasions, we ailed o ge his da a because he Silkki ie se e was o line. In addi ion, we migh ha e missed some in o ma ion i a selle emo ed a p oduc om he ma ke place be ween he measu emen s. Simila ly, i is possible ha an old p oduc could be egis e ed as new because he selle changed he p oduc i le. We belie e, howe e , ha hese limi a ions a e no c i ical o ou main analyses, which concen a ed on selle epu a ion and capaci y. Fu he mo e, ou da a is ex ensi e, so he missing da a is unlikely o ha e an impac on he analyses. To ou knowledge, ou da a se is one o he la ges one ha has been collec ed om he c yp oma ke s. Fu he mo e, ou da a se is no ulne able o he challenges conce ning he collec ion o online da a om o he pla o ms, such as he Silk Road. In he Silk Road, o example, incomple e c awls may lead o misleading esul s (Munksgaa d e al., 2016). 8 In his s udy, we we e able o ollow he ex en o ading and p oduce almos eal- ime s a is ics abou how people buy and sell illegal d ugs. Ou me hod is highly eliable, and he long pe iod o da a collec ion gua an ees his eliabili y. Hence, his s udy con ibu es o he inc easing body o li e a u e s udying c yp oma ke s wi h web c awle s (e.g., Ald idge, 2016; Al idge and Déca y-Hé u, 2016; Ch is in, 2013; Deman e al., 2016; Ha dy and No gaa d, 2015). Mos impo an , ou esul s help e eal he consume in e ac ions be ween selle s and buye s on online c yp oma ke s, and ou indings unde line he ele ance o bo h epu a ion and capaci y as pa s o he c yp oma ke us sys em. Fu u e s udies should con inue o collec longi udinal da a se s. 5. Re e ences Ald idge, J., Déca y-Hé u, D., 2016. Hidden wholesale: he d ug di using capaci y o online d ug c yp oma ke s. In . J. D ug Policy. 35:7-15. Allison, P. Bewa e o so wa e o ixed e ec s nega i e binomial eg ession [A icle on he In e ne ]. Ci y: Publishe ; Yea [upda ed 8 June 2012; ci ed 4 Augus 2016]. A ailable om: h p://s a is icalho izons.com/ e-nb eg Banc o , A., Reid, P.S., 2017. Challenging he echno-poli ics o anonymi y: he case o c yp oma ke use s. In . Commun. Soc. 20:497-512. Ba a , M.J., Fe is, J.A., Wins ock, A.R., 2014. Use o Silk Road, he online d ug ma ke place, in he Uni ed Kingdom, Aus alia and he Uni ed S a es. Addic ion. 109(5):774-83. Ba a , M.J., Fe is, J.A., Wins ock, A.R., 2016a. Sa e sco ing? c yp oma ke s, social supply and d ug ma ke iolence. In . J. D ug Policy. 35:24–31. Ba a , M.J., Len on, S., Maddox, A., Allen, M. 2016b. ‘Wha i you li e on op o a bake y and you like cakes?’—D ug use and ha m ajec o ies be o e, du ing and a e he eme gence o Silk Road. In . J. D ug Policy. 35:50-57. Ba a , M. J., Maddox, A. 2016. Ac i e engagemen wi h s igma ised communi ies h ough digi al e hnog aphy. Quali . Res. 16:701-19. Bohannon, J., 2016. Bi coin bus s. Science. 351:1144-6. B oséus, J., Rhumo ba be, D., Mi eaul , C., Ouelle e, V., C ispino, F., Déca y-Hé u, D. 2016. S udying illici d ug a icking on Da kne ma ke s: s uc u e and o ganisa ion om a Canadian pe spec i e. Fo ensic Science In e na ional 264:7-14. Caude illa, F., Ven u a, M., Fo nís, I., Ba a , M. J., Vidal, C., Quin ana, P., ... Calzada, N. (2016). Resul s o an in e na ional d ug es ing se ice o c yp oma ke use s. In . J. D ug Policy. 35:38-41. Ch is in, N., 2013. T a eling he Silk Road: a measu emen analysis o a la ge anonymous online ma ke place. In: WWW ’13 P oceedings o he 22nd In e na ional Con e ence on Wo ld Wide Web, pp. 213–224. New Yo k: ACM. 9 Déca y-Hé u, D., Giommoni, L. (2017). Do police c ackdowns dis up d ug c yp oma ke s? A longi udinal analysis o he e ec s o Ope a ion Onymous. C ime, Law and Social Change 67:55-75. Déca y-Hé u, D., Leppänen, A., 2013. C iminals and signals: An assessmen o c iminal pe o mance in he ca ding unde wo ld. Secu i y Jou nal. 29:442. Deman , J., Munksgaa d, R., Houbo g, E. 2016. Pe sonal use, social supply o edis ibu ion? C yp oma ke demand on Silk Road 2 and Ago a. T ends in O ganized C ime, doi:10.1007/s12117-016- 9281-4 Dingledine, R., Ma hewson, N., Sy e son, P., 2004. To : he second-gene a ion onion ou e . Washing on, DC: Na al Resea ch Lab. Du , C., 2005. Pa y d ugs and pa y people: examining he ‘no maliza ion’ o ec ea ional d ug use in Melbou ne, Aus alia. In . J. D ug Policy. 16:161-70. Dolli e , D.S., 2015 E alua ing d ug a icking on he To Ne wo k: Silk Road 2, he sequel. In J D ug Policy. 26:1113-23. Dolli e , D.S., Kenney, J.L., 2016. Cha ac e is ics o D ug Vendo s on he To Ne wo k: A C yp oma ke Compa ison. Vic . O ende s. May 2, DOI: 10.1080/15564886.2016.1173158 Dolli e , D.S., Kuhns, J.B. 2016. The P esence o New Psychoac i e Subs ances in a To Ne wo k Ma ke place En i onmen , J. Psychoac . D ugs. 48: 321-329, DOI: 10.1080/02791072.2016.1229877 Eu opean Moni o ing Cen e o D ugs and D ug Addic ion. Finland coun y o e iew. Lisbon: EMCDDA; [ci ed 4 Augus 2016]. A ailable om: h p://www.emcdda.eu opa.eu/coun ies/ inland FBI, 2014. Sealed Complain agains Ross Ulb ich . Na co ic T a icking Conspi acy. h ps://www.documen cloud.o g/documen s/801103-172770276-ulb ich -c iminal-complain .h ml Finnish Cus oms (2016). P e-in es iga ion epo s o he Douppikauppa Case. 9010/R/6369/15; 9010/R/7463/15; 9010/R/2757/16; 9010/R/3004/16. Tampe e: Finnish Cus oms. Glassman, M., Kang, M.J., 2012. In elligence in he in e ne age: he eme gence and e olu ion o open sou ce in elligence (OSINT). Comp. Hum. Beha . 28:673-82. Hakka ainen, P., Kainulainen, H., Pe älä, J. 2006. Suomen kannabisma kkina – paljonko pil eä palaa? [Measu ing he size o he cannabis ma ke in Finland]. Yh eiskun apoli iikka. 71: 583-95. Hakka ainen, P., Ka jalainen, K., Ojajä i, A., Salasuo, M., 2015. Huumausaineiden ja kun odopingin käy ö ja nii ä koske a mielipi ee Suomessa uonna 2014 [D ug use, doping and public opinion in Finland: esul s om he 2014 D ug Su ey]. Yh eiskun apoli iikka. 80(4):319-33. Finnish. Ha dy, R.A., No gaa d, J.R., 2015. Repu a ion in he In e ne black ma ke : an empi ical and heo e ical analysis o he Deep Web. J. Ins . Econ. 12:515–39. House , D., Woode s, J. 2006. Repu a ion in auc ions: Theo y, and e idence om eBay. Jou nal o Economics & Managemen S a egy. 15:353–69. 16 Figu e 3: Change in daily sales on Silkki ie based on pe cen ile alues o epu a ion and a ailabili y (adjus ed p edic ions based on ixed-e ec s eg ession models) 62 75 89 101 112 126 142 173 222 97 107 115 122 128 136 143 153 167 0 50 100 150 200 250 10 20 30 40 50 60 70 80 90 Eu os Pe cen iles Repu a ion Capaci y