Driving licenses based on points systems: Efficient road safety strategy or latest fashion in global transport policy? A worldwide meta-analysis

D i ing licenses based on poin s sys ems: E ficien oad sa e y s a egy o la es ashion in global anspo policy? A wo ldwide me a-analysis Jose ´I. Cas illo-Manzano n , Me cedes Cas o-Nun ˜o Facul ad de Ciencias Econo ´micas y Emp esa iales, Uni e si y o Se ille (Spain), A da. Ramo ´n y Cajal, 1, 41018 Se ille, Spain a icle in o Keywo ds: Road sa e y D i ing license Poin - eco d sys ems Me a-analysis app oach abs ac One o he mos popula coac i e measu es de eloped o p e en oad a fic acciden s in ecen decades is he implemen a ion o d i ing licenses based on poin s sys ems (PS) which penalize epea o ende s wi h suspension o wi hd awal o hei licenses. This pape analyzes hei apid sp ead wo ldwide h ough an in-dep h e iew o he exis ing li e a u e. A comp ehensi e me a-analysis o he e ec s o PS on oad a fic acciden s and he du a ion o hese e ec s has been conduc ed. The findings show ha he s ong ini ial posi i e impac (15 o 20% educ ions in acciden s, a ali ies and inju ies) seems o wea o in unde eigh een mon hs. This limi ed e ec i eness is ela ed o he absence o complemen a y en o cemen o back up hese measu es. Wi hou hem, poin s sys ems could u n in o a boome ang oad sa e y policy, and e en be abandoned a a la e da e. The implica ions o he conclusions o legisla ion and u u e esea ch a e conside ed. &2012 Else ie L d. All igh s ese ed. 1. In oduc ion The ex ao dina y economic g ow h expe ienced h oughou he 20 h cen u y has aised a es o mo o ehicle owne ship all a ound he wo ld while, a he same ime, oad a fic acciden s (RTA) ha e become one o he le hal epidemics a ec ing human- kind, wi h 1.2 million people killed and up o 50 million mo e inju ed e e y yea (WHO, 2009). This is, howe e , a silen kille which o yea s has been he objec o much less media a en ion han o he , less e e yday bu mo e eye-ca ching agedies, such as ai acciden s o na u al disas e s. Acco ding o he same sou ce, unless a p e en a i e s a egy is adop ed on a global scale, RTA will become he fi h leading cause o dea h wo ldwide by 2030. In many coun ies, he implemen a- ion o inc easingly se e e oad sa e y policies, oge he wi h he ac ha he cu en economic c isis has educed a fic olumes (a leas in he sho e m), ha e p oduced a la ge all in he main a fic acciden indica o s du ing he fi s decade o he 21s cen u y (see OECD, 2010a,2010b). Howe e , his imp o emen has no been uni o mly dis ibu ed in geog aphical e ms, wi h huge di e ences be ween high-income coun ies and middle- income/low-income coun ies (IRF, 2009). As hey ha e achie ed a mo e-o -less es ed minimum deg ee o e ec i eness (WHO, 2009), he majo i y o he in e en ions in he de eloped coun- ies ha e been p og essi ely imi a ed in de eloping coun ies. We ound basically wo g oups o ac ions: fi s ly, s a egies ha ocus on elemen s ha a e ex e nal o he oad use (such as in as uc u e imp o emen s—analyzed by A ´l a ez e al., 2007 o sa e ehicle design—s udied by Ch is ensen and El ik, 2007); and secondly, s a egies ha ocus on p e en ing o co ec ing unsa e oad-use beha io s, based on he e idence ha he human elemen is he ac o ha explains he majo i y o acciden s (S an on and Salmon, 2009). This las block o ac ions, which has been g ea ly de eloped in he academic li e a u e, can in u n be classified as: p e en i e measu es ha seek o pe suade o dissuade h ough educa ion and in o ma ion (such as communica ion campaigns and oad sa e y ad e isemen s, conside ed by Cas illo-Manzano e al., 2012 and Lewis e al., 2008); o co ec i e o sanc ioning measu es (mone a y and non-mone a y) which y o in e nalize he social cos s o RTA h ough coac i e dissuasion (Cas illo-Manzano e al., 2011). Apa om he applica ion o se e e c iminal penal ies by Law (judicial sen ences and imp isonmen ) o punish he se e es in ingemen s, o he so e egula ions o mino in ingemen s based on economic sanc ions (by fines and insu ance paymen s) and license dep i a ion (by suspension o wi hd awal) also s and ou (Bou geon and Pica d, 2007). In oad a fic legisla ion in many coun ies, license dep i a- ions ha e e ol ed in o a ype o d i ing license based on a poin s sys ems (hence o h PS) which, despi e di e ences be ween coun- ies, espond o a common philosophy ha can be summa ized ollowing Nole ´nandO ¨s lin, (2008): se e al specific o enses com- mi ed by d i e s cause he addi ion/loss o a ce ain numbe o poin s and when he numbe o poin s eaches he maximum allowed o all poin s a e los , he d i e ’s license is suspended. Con en s lis s a ailable a SciVe se ScienceDi ec jou nal homepage: www.else ie .com/loca e/ anpol T anspo Policy 0967-070X/$ - see on ma e &2012 Else ie L d. All igh s ese ed. doi:10.1016/j. anpol.2012.02.003 n Co esponding au ho . Tel.: þ34 954 556727; ax: þ34 954 557629. E-mail add esses: [email p o ec ed] (J.I. Cas illo-Manzano), [email p o ec ed] (M. Cas o-Nun ˜o). T anspo Policy 21 (2012) 191–201 Despi e hei widesp ead use in p ac ice, scien ific s udies on he e ec i eness o PS a e no de eloped me hodologically in he academic li e a u e o he same deg ee, and he e is a ce ain amoun o con o e sy ega ding he du a ion o hei e ec s (Cas illo-Manzano e al., 2010). The e a e some s udies ha a e mo e op imis ic abou he impac and sus ainabili y o he measu e’s posi i e e ec s o e ime, including con ibu ions o he success o o he oad sa e y measu es (such as Zambon, e al. 2008 which a ibu e o i a majo inc ease in he wea ing o sea bel s); while o he au ho s, such as Mon ag (2010), claim ha i s impac is only sho - e m and ha , pa adoxically, he e migh e en be some pe e se e ec s, ei he because a fic law en o ce- men p oduces mo e isk- aking beha io s, o because he e is a dec ease in he in ensi y o police ac i i ies due o us in he de e en e ec o he sys em on i s own. 1 In his line, PS can encou age undesi able beha io s, such as hi -and- un c ashes (SWOV, 2010) and/o d i ing wi hou a alid license (Knox, e al., 2003). In ac , ollowing Engs ¨ om e al., (2003), e y li le is known abou he e ec s o he PS on oad sa e y, because as SWOV (2010) s a es, i is e y di ficul o isola e he impac ha hese sys ems ha e om he e ec s o o he complemen a y ypes o en o cemen being applied a he same ime (i.e., an inc ease in police o law en o cemen , public campaigns). The aim o his pape is o discuss some ea u es o he poin s- eco d mechanism, while simul aneously de eloping an empi ical amewo k o in es iga e in mo e de ail i his eally is an e ficien global s a egy o imp o ing oad sa e y o simply a ashion policy which is imi a ed om one coun y o ano he . We use a me a-analysis app oach o in es iga e he e ec i eness and sho - and long- e m e ec s o his sys em, de eloped in he amewo k o a fixed-e ec s model and a andom-e ec s model. Ra he han jus a simple s a e-o - he-a li e a u e e iew, we selec academic esea ch om a wide sea ch o s udies de eloped a ound he wo ld conside ed as ele an on he basis o he me hodologies used and empi ical findings, and we summa ize a esul . This unde s anding is conside ed o be especially use ul o u u e expansion o he policy o de eloping o unde eloped coun ies as i p o ides he keys equi ed o i s op imum design. The pape is o ganized as ollows: ollowing his in oduc ion, Sec ion 2 examines d i ing licenses based on PS as an in e na ional s a egy o educing a fic acciden s. In Sec ion 3,weexplain he design o he me a-analysis model applied, and discuss he findings. In Sec ion 4 he main conclusions a e d awn and he implica ions o oad sa e y ha de i e om his esea ch a e se ou . 2. Poin s sys ems o imp o e oad sa e y D i ing licenses based on a poin s- eco d ha e hei p ede- cesso s in he license endo semen sys ems ha we e in oduced in he nine een- hi ies (Milulik, 2007). They ha e cu en ly been ex apola ed o o he a eas o public poli ics no only ela ed o he anspo a ion sec o ( o in ac ions in fishe ies ac i i y in he Eu opean Union; o illegal immig a ion in coun ies such as Canada, Aus alia and New Zealand; o s ic codes o discipline a high schools in Hong Kong and India). In he p esen day, his oad sa e y s a egy has been pa icula ly ecommended by in e na ional ins i u ions because o i s po en ial, a leas in he ini ial pe iod a e i s in oduc ion, (see Global Road Sa e y Pa ne ship, 2008;SWOV, 2010). I has p og essi ely gained popula i y and accep ance in public opinion wo ldwide (Nole ´n and O ¨s lin, 2008) because i is conside ed o be mo e jus when dealing mo e s ic ly wi h epea o ende s and ai e han a mone a y fine, ( he eal e ec o which depends on he o ende ’s pu chasing powe , ETSC, 2008). Acco ding o El ik and Vaa (2004), his measu e o ms pa o a wide oad sa e y s a egy ha includes wo o he en o cemen measu es: wa ning le e s 2 and d i ing license e oca ions. I is also an umb ella s a egy, as i co e s all ypes o a fic iola ions compa ed o o he measu es included in he gene al legisla i e sys em which a e exclusi ely ela ed o one specific o ense a ea (speeding, alcohol). A e iew o he PS in di e en coun ies e eals ha , al hough hei s uc u es a e all simila , he e a e di e ences in hei designs and applica ion (Nole ´n and O ¨s lin, 2008). F om hei analysis i can be concluded ha hey a y conside ably in he selec ion o o enses and hei sco ing, he ways o coun ing poin s, he h esholds o disqualifica ion o du a ion o disqua- lifica ion and adminis a ion p ocedu es. E en in he same coun- y di e ences can be ound be ween s a es as, in ce ain coun ies, hese a e sys ems ha a e ans e ed o he au ho i ies o he a ious ju isdic ions. This is he case o he Uni ed S a es, Aus alia and Canada. Fu he mo e, he punishmen applied by PS o a a fic iola ion may no be s anda dized, as in coun ies such as I eland and he Uni ed Kingdom he numbe o poin s can ise o he same iola ion depending on whe he he d i e is ound guil y in cou . The penaliza ion is he e o e adjus ed o each iola ion depending on i s se iousness o whe he i is a epe i ion o he same iola ion (ETSC, 2008). 3 The e a e wo equi alen a ia ions wi h di e en mechan- isms used o compu e he poin s ha penalize o enses (see, Nole ´n and O ¨s lin, 2008): (1) Penal y PS (PPS) o poin deduc ion: all d i e s who possess a alid license a e alloca ed an ini ial numbe o poin s om which poin s a e deduc ed depending on he o ense commi ed. Once his ini ial c edi uns ou he license is wi hd awn. ‘‘The poin is o gain no poin ’’ (ETSC, 2008); (2) Deme i PS (DPS) o poin accumula ion: all d i e s who possess a alid license s a om sc a ch, i.e., no poin s, and accumula e poin s, depending on he o ense commi ed. The license is wi h- d awn a e a ce ain numbe o poin s a e accumula ed. In bo h cases, he deduc ion o appo ioning o poin s is o a pe sonal na u e, ha is, i is di ec ly linked o he d i ing license owned by he o ende so ha when she/he is caugh by he police commi ing a a fic iola ion he poin s canno be used agains he ehicle owne ia he ehicle egis a ion numbe (SWOV, 2010). Simila ly, in bo h o he e sions, all d i e s s a in he same si ua ion (maximum poin s/ ze o poin s), al hough in some coun ies like La ia o he Uni ed Kingdom, di e en ial sys ems a e in place o g oups o d i e s wi h special ci cums ances, such as no ice d i e s (wi h a s ic e sys em) 4 ; o p o essional d i e s (wi h a mo e pe missi e sys em). Acco ding o Bou geon and Pica d (2007), in some cases such as F ance o Spain, ‘‘ edemp i e sys ems’’ a e applied ha a e no 1 Acco ding o Dionne e al., (2011),‘‘y he police a e he mos impo an en o ce s o he di e en incen i e schemes’’. 2 Gene ally-speaking, ‘‘wa ning le e s’’ a e an incen i e sys em used o modi y negligen d i ing beha io in oad sa e y (Wilde and Mu doch, 1982) when a d i e commi s a ce ain a fic o ense (e.g., exceeding he speed limi , o in pho o ada en o cemen ). Howe e , in some coun ies o US s a es wi h poin s sys ems, like F ance, Michigan (U.S.), On a io (Canada) and Spain, wa ning le e s a e also used as ‘‘ad iso y w i en wa nings’’ which a e sen o d i e s who a e beginning o app oach he limi o license wi hd awal (see, o example Cas illo- Manzano e al., 2010 and Haque, 1990). 3 This modali y, which consis s o g adua ing he penaliza ion applied o each iola ion depending on i s se iousness, is no en i ely sanc ioned by expe s such as El ik and Vaa (2004), who belie e i is no e y eliable due o i s subjec i i y. 4 PS can be specifically a ge ed a no ice d i e s (Ne he lands and Finland) o unde G adual Licenses (GDL) and p oba iona y sys ems, which consis o a fic iola ions du ing an in e media e s age o licensing possibly delaying he no ice d i e a aining a ull license (e.g., a new G adua ed Deme i Poin Sys em o no ice d i e s in Aus alia, since Decembe , 2010). J.I. Cas illo-Manzano, M. Cas o-Nun ˜o / T anspo Policy 21 (2012) 191–201192 only o a puni i e na u e bu also ha e a ehabili a i e aspec o hem. In hese sys ems he d i e who has been penalized can ha e he poin s she/he has accumula ed and emo ed, o eco e hose ha ha e been los , once a ce ain pe iod o ime has passed wi h no u he o enses being commi ed, a e a ending a e- educa ion and oad-sa e y awa eness cou se sel -financed by he o ende , and e en a e e aking he d i ing es o eco e he license. On occasion, ex a poin s a e added o deduc ed om he licenses o good d i e s who do no commi any o ense du ing a ce ain pe iod o ime (Nole ´nandO ¨s lin, 2008); his is he case in Malaysia, whe e he so-called KEJARA Sys em is applied, and Spain. In ju isdic ions such as he s a e o New Sou h Wales (Aus alia), he ‘‘double penal y poin s’’ sys em is used o inc ease he le el o penaliza ion o ce ain o enses (such as speeding) du ing holiday pe iods (highe poin s du ing high- isk pe iods). In some s a es in he US and Eu opean coun ies like Ge many and I eland, PS can also p oduce o he ex e nal e ec s: insu ance companies equi e d i e s’ poin s- eco ds o be epo ed o hem o de e mine whe he o no o enew an insu ance policy (ETSC, 2008 is analyzing he opic). This is an a emp o c ea e a s onge incen i e o eckless d i e s who do no espond o ha o insu ance p icing (Dionne e al., 2011). I espec i e o he pa icula ea u es o he PS a ian used, Basili and Nici a (2005) highligh he ac ha he measu e also aspi es o igge a numbe o a o able e ec s in d i e beha io : 1. De e ence o he abili y o make o ende s associa e epea a fic iola ions wi h mo e se e e sanc ions o he poin ha hey lose hei licenses (Zambon e al., 2008). I could be said ha he epea ed iola ion o a fic egula ions is p oducing a g owing ma ginal cos o he o ende ; 2. Selec ion o emo al o he mass o d i e s who epea edly iola e a fic egula ions wi h he objec i e o educing he likelihood o a fic acciden s (Diaman opoulou e al., 1997); 3. Co ec ion o incen i e o d i e s o ec i y hei inapp op ia e beha io a he wheel (Poli de Figuei edo e al., 2001); 4. Educa ion o P e en ion, as d i e s a e shown which a e he mos se ious punishable iola ions and mos dange ous beha io s whils (in some coun ies) i is a he same ime es ablished ha cou ses on oad sa e y ha e o be a ended (Roca and To osa, 2008). In Table 1, we use he agg ega ion o coun ies in egions p oposed by WHO (2009) as an ini ial s udy c i e ion o dis in- guish be ween coun ies wi h PS (44) whils also indica ing which a ian has been implemen ed, PPS o DPS. The hi d column o his Table also p o ides a sample o coun ies om each WHO egion o which some kind o s udy has been ound on he e ec s o he PS, i espec i e o he me hodology used o i s measu e- men , and oge he wi h i s implemen a ion da e. De ailed analysis o Table 1 has allowed a se ies o findings o be de e mined:  The a eas whe e he de elopmen o he PS has been mos widesp ead can be seen o be hose which include mo e high- income coun ies (HIC), especially in he Eu opean Region. This sugges s ha coun ies need o ha e he app op ia e echnical means o implemen he sys em, as well as a ce ain deg ee o oad sa e y cul u e (as desc ibed by Dodge and Ki chin, 2007). Fu he mo e, he sys em ha is mos adop ed in he majo i y o he coun ies conside ed is he DPS me hod, i.e., he o iginal model.  Twen y- ou coun ies ha e op ed o he sys em in he Eu opean Region. This measu e has become a gene alized s anda d o oad sa e y in wes e n coun ies, whe eas i s applica ion in Eas e n Eu opean coun ies has been much mo e limi ed and mos ly confined o coun ies ha a e in he EU, such as he Czech Republic o La ia, amongs o he s, and also Tu key, which has applied o join. I seems clea ha belonging o he EU and i s a ea o influence has been he h ead ha has led coun ies o imi a e one ano he . Indeed, i could be said ha a e he Eu opean Commission se a a ge o a 50% educ ion in a fic a ali ies be ween 2000 and 2010 (Eu opean Commission, 2001), i was no so much an imi a ion e ec ha could be spoken o among he g oup o measu es ha Eu opean coun ies adop ed, bu a he a con agion e ec . 5 The e a e also some EU coun ies whe e d i ing licenses a e no in ac based on he PS, al hough hey do bea many simila i ies. One example is Po ugal, whe e a d i ing license can be wi hd awn i a cou decides he o ende has been guil y o e y se ious o enses h ee imes, o o se ious o enses fi e imes, in h ee yea s, and, u he mo e, d i e s mus e ake hei d i ing es s o ge hei licenses back.  Ch onologically, a e i s pionee ing in oduc ion in he US o e fi y yea s ago, he sys em sp ead o Asia and Aus alia and he ea e o a Eu opean coun y (Ge many). I has only eally expanded and been pe ec ed in ecen imes as i was in he las decade ha he g ea es numbe o changes and con e s we e seen, especially in Eu ope. The la es egions o in oduce PS ha e been A ica and he Eas e n Medi e anean.  Finally, we find cul u al simila i ies in he PS a ian adop ed. A e fi s being in oduced in he US, he DP sys em has been eplica ed in o he Anglo–Saxon coun ies i espec i e o he WHO egion o which hey belong (Aus alia, New Zealand, Sou h A ica, Canada, UK, and I eland o hei o me colonies, such as Hong Kong). In mo e ecen imes, i has been he PPS ha has been implemen ed in La in Eu opean coun ies (F ance, I aly and Spain), and o la e in hese coun ies’ o me Ame ican colonies, such as Ecuado and A gen ina, al hough no in colonies ha come unde he cul u al influence o he US, like Mexico and Panama, which use he DPS. Resea che s such as P iya and U eng (2009) find co ela ions be ween cul u al elemen s and acciden s on he basis o li es yle and socioeconomic aspec s. These s udies would uphold he idea ha he in e na ional sp ead o his oad sa e y s a egy is no only explained by he imi a ion e ec ha a ises be ween coun ies because o hei geog aphical p oxi- mi y, bu also because o cul u al issues, as i would appea ha he a ian ha is applied is closely linked o he peculia i ies o each coun y. 3. Me a-analysis o e ec i eness o poin sys ems mechanisms One o he objec i es o his s udy is o syn hesize a se ies o esul s on he size and du a ion o PS e ec s wo ldwide om p e ious s udies ha ha e used a ious me hodologies. The me a- analysis app oach has ad an ages o e a li e a u e e iew o his because o i s s a is ical na u e and i s in e ence p ocess (Glass e al., 1981). This app oach has been b oadly de eloped o e he las 20 yea s, basically in expe imen al sciences and en i onmen al eco- nomics, al hough a a ie y o p eceden s can also be ound o i s being applied o a ange o oad sa e y s a egies in ecen yea s (see o example E ke, 2009, o ed-ligh came as; Høye, 2010, o on ai bags; o Phillips, e al., 2011, o ad e ising campaigns). As a as PS a e conce ned, he e is a p eceden in El ik e al. (2009),who syn hesized a numbe o specific s udies ela ed o DPS-linked elemen s and license suspensions (wa ning le e s, special d i ing es s and imp o emen cou ses). 5 As has been seen om he measu es ha ha e been p og essi ely applied by he a ious coun ies, acco ding o OECD, (IRTAD) (2010). J.I. Cas illo-Manzano, M. Cas o-Nun ˜o / T anspo Policy 21 (2012) 191–201 193 3.1. S udy design: da a coding and a iables An ini ial sea ch was conduc ed in he academic li e a u e o iden i y he mos ecen esea ch on he opic o he e ec s o PS a ound he wo ld, wi h he cons ain o he publica ion o m (a icles in pee e iewed scien ific jou nals o ins i u ional epo s). 6 Using gene al sea ch e ms in o de ha no s udy should be excluded ap io i, hese s udies ha e been ound in he ollowing sou ces: he epu ed T anspo a ion Resea ch In o ma ion Se ices (TRIS) da abase (U.S. Depa men o T anspo a ion), PubMed (online da abase o he U.S. Na ional Lib a y o Medicine), he ISI Web o Knowledge, Scopus and Science di ec (Else ie s online da abases), epo s om esea ch ins i u es such as SWOV (The Ne he lands) and TRL (UK) and In e ne sea ches by Google Schola . In his way almos fi y s udies wi h hese cha ac e is ics we e loca ed al hough hey a e all g ea ly he e ogeneous, especially wi h ega d o hei me hodologies. Some a e eminen ly heo e ical (such as Basili and Nici a, 2005 and Bou geon and Pica d, 2007), whils o he s a e based on a compa ison o be o e-and-a e s a is ics (Poli de Figue eido e al., 2001). Only a sample o sui able s udies was selec ed, using ypical inclusion c i e ia in me a-analysis (see El ik, 1999), i.e., hose which: 1. Include an es ima ion o PS e ec i eness and/o du a ion using s a is ical o econome ic me hodology ha seeks o sepa a e ou he e ec s o PS om o he a iables ha migh ha e had an e ec du ing he pe iod analyzed (bo h me a-analyses and s udies ha only e alua e elemen s such as wa ning le e s and e- educa ion cou ses we e ejec ed since, as was indica ed in Sec ion 2, hese a e no applied equally in all coun ies); 2. Repo he s anda d de ia ion, confidence in e al, sample size o simila s a is ical in o ma ion abou he accu acy and obus ness o he esul s. The ini ial da a sample was efined by he applica ion o hese quali y indica ions and a o al o 26 s udies om a ound he wo ld we e finally included in he me a-analysis, many o hem wi h mul iple ele an esul s which could be used o expand he final sample. This sui able me hod o da a collec ion was cha - ac e ized by a coding p ocedu e, as seen in Tables 2–5. Following he me hodology desc ibed in Sec ion 3.2, below, wo scena ios we e en isaged o he me a-analysis: SCENARIO (I)Es ima ion o summa y e ec s o PS: To minimize he likely he e ogenei y among he 26 sui able s udies, hese we e classified in o 3 subg oups depending on he ype o a iables ha we e conside ed o e alua ing he e ec s o he PS. As he e a e s udies wi h mul iple esul s, he sample size will be he o al numbe o esul s compu ed 7 :  I VTRA: Va iables ela ed di ec ly o RTA, in e ms o educ- ions in he numbe o acciden s, casual ies, a ali ies o inju ed. These a iables we e ound in 13 s udies wi h 24 esul s (m¼24). Table 1 Coun ies wi h poin s sys ems in he who egions. Sou ce: Au ho s. based on WHO (2009), ETSC (2008) and oad a fic legisla ion in each coun y. WHO egion No. o coun ies/ income le el Coun ies wi h a poin s sys em d i ing license (DPS o PPS) such asySelec ed coun ies/yea poin s sys em implemen ed 141 (0 HIC, 11 MIC, 30 LIC) Sou h A ica (DPS) Sou h A ica (in P e o ia, 2006) 232 (6 HIC, 26 MIC, 0 LIC) A gen ina (PPS), Be muda (DPS), B azil (DPS), Canada (DPS), Ecuado (PPS), Jamaica (DPS), Mexico (DPS), Panama (DPS), Pe u (DPS), Uni ed S a es o Ame ica (DPS) – USA (s a e by s a e, i.e. Connec icu , 1957) – Canada (s a e by S a e, i.e. Quebec, 1978) – B azil (1998) 310 (0 HIC, 6 MIC, 4 LIC) – – 420 (5 HIC, 12 MIC, 3 LIC) Qa a (DPS), The Uni ed A ab Emi a es (DPS) The Uni ed A ab Emi a es (2008) 549 (25 HIC, 21 MIC, 3 LIC) Aus ia (DPS), Bulga ia (DPS), Cyp us (DPS), The Czech Republic (DPS), Denma k (DPS), F ance (PPS), Ge many (DPS), G eece (PPS), Hunga y (DPS), Iceland (DPS), I eland (DPS), Is ael (DPS), I aly (PPS), La ia (DPS), Li huania (DPS), Luxembou g (DPS), Mal a (DPS), No way (DPS), Poland (DPS), Romania (DPS), Slo enia (DPS), Spain (PPS), Tu key (DPS), Uni ed Kingdom (DPS) – Ge many (1974) – Uni ed Kingdom (1988) – F ance (1992) – Finland (1996) a – I eland (2002) – The Ne he lands (2002) a – I aly (2003) – Aus ia (2005) – Spain (2006) – The Czech Republic (2006) 626 (6 HIC, 15 MIC, 5 LIC) Aus alia (DPS), China (DPS), Japan (DPS), Malaysia (DPS), New Zealand (DPS), Republic o Ko ea (DPS), Singapo e (DPS), – New Zealand (1967) – Japan (1968) – Aus alia (s a e by s a e, i.e. Vic o ia, 1970) – China (ju isdic ion by ju isdic ion, i.e. Hong Kong, 1984) No es: DPS: Deme i poin sys em, PPS: Penal y poin sys em. HIC: High-income coun ies; MIC: Middle-income coun ies; LIC: Low-income coun ies. 1. A ican Region; 2. Region o he Ame icas; 3. Sou h-Eas Asia Region; 4. Eas e n-Medi e anean Region; 5. Eu opean Region; 6. Wes e n Pacific Region. a Only Poin s Sys ems o no ice d i e s on p oba ion. 6 In o de o educe any possible bias, sou ces and bibliog aphy we e chosen blindly by he au ho s o he s udy. 7 Following El ik and Vaa, (2004) we assume he independence hypo hesis be ween mul iple esul s. J.I. Cas illo-Manzano, M. Cas o-Nun ˜o / T anspo Policy 21 (2012) 191–201194  I VDB: Va iables ela ed o d i e beha io , in e ms o educ ions in isky o dange ous conduc , d i ing less unsa ely (a smalle pe cen age o d i e s no wea ing sea bel s, o example) o a educ ion in law iola ions (speed limi s, d ink- d i ing). These a iables we e ound in 8 s udies wi h 18 esul s (m¼18).  I VHCD: Va iables ela ed o heal h ca e da a, in e ms o educ ions in RTA ela ed A&E admissions, su ge y o hospi- aliza ions. These a iables we e ound in 5 s udies wi h 8 esul s (m¼8). SCENARIO (II)Es ima ion o he du a ion o he PS e ec : The a iable conside ed in his case (LENGTH OF EFFECTS) is he du a ion o PS e ec s, in numbe o mon hs, es ima ed o each o he s udies. As can be seen in Tables 2–5, no all he sui able s udies analyze his issue. To be p ecise, we shall wo k wi h he sample o 20 s udies ha do (m¼20). 3.2. Me hodological aspec s. The es ima ion me hod used was he ‘‘log-odds model’’ ound in El ik and Vaa (2004). In his me hod, he es ima ions o he sui able s udies a e combined o ob ain a summa y-measu e using mean weigh ing acco ding o he s a is ical weigh ( he in e se o he p ecision o he s udy). Acco ding o he ollowing o mula, he mean weigh ed e ec o ‘‘m’’ es ima ions ðYÞwill be: Y¼exp P m i¼1 w i y i P m i¼1 w i  ð1Þ When: y i ¼log (i- h es ima ion ob ained by each s udy). w i ¼ he s a is ical weigh o he i- h es ima ion ob ained as w i ¼ð1= i Þ,when i is he a iance o each i- h es ima ion, whe eby P m i¼1 w i ¼1: The possible he e ogenei y o he s udies can be aken in o accoun using he andom-e ec s model (REM), o no included i he fixed-e ec s model (FEM) is used. The FEM assumes ha he e is one single e ec on he popula ion and does no ake in o accoun he a iabili y o he esul s among he di e en s udies. The size o he s udy and i s own a iance (in a-s udy a iabili y) a e he only de e minan s o i s weigh in he me a-analysis. The De simonian and Lai d Q s a is ic (1986) is calcula ed o e i y he alidi y o his hypo hesis, dis ibu ed in acco dance wi h a Chi- Squa e wi h m-1 deg ees o eedom (m¼no. o combined es ima ions): Q¼X m i¼1 w i y 2 i ðP m i¼1 w i y i Þ 2 P m i¼1 w i ð2Þ I he null hypo hesis is ejec ed (H 0 : non he e ogenei y) i is ecommended ha REM be used as his does ake in o accoun Table 2 Selec ed s udies in he Ame icas egion: coded da a. Sou ce: Au ho s. Coun y/ income le el Resea che /place P oxy o e ec s Me hodology Resul s TIME se ies s udied/– ime– span o e ec s B azil (MIC) Libe a i e al. (2001)/ (ci y o Lond ina) VTRA Be o e/ A e analysis using Chi-squa ed and Fishe ’s es s. A educ ion o 20% o ca and 9.1% o mo o cycle acciden ic ims. An inc ease o 39% in use o sa e y bel s and o 112% in use o helme . (1997–1998)/6 mon hsVBD Ma ei de And ade e al. (2008)/(ci y o Lond ina) VTRA Linea eg ession. A 28.4% g ea e educ ion in mo ali y le els han unde o he p e ious measu es. (2000–2005)/12 mon hs Canada (HIC) Chen e al. (1995)/ (B i ish Columbia) VTRA Logis ic eg ession. The esul s showed a consis en inc ease in pos pe iod acciden s pe d i e wi h inc easing p e pe iod numbe o c ashes. (1985–1990)/24 mon hs Dionne e al. (2011)/ (Quebec) VBD P opo ional Haza ds Model. The equency o a fic iola ions was educed by 15% and he isk o a RTA a e a a fic o ense was 20% lowe . (1985–1996)/12 mon hs Redelmeie e al. (2003)/(On a io) VBD Case-c osso e s udy The isk o a a al c ash in he mon h a e a con ic ion was abou 35% lowe . (1988–1999)/2 mon hs Uni ed S a es o Ame ica (HIC) Gebe s and Peck (2003)/(Cali o nia) VBD Bo h mul iple linea eg ession analysis and canonical co ela ion. An imp o emen o 14.9% in classifica ion o p edic ing and iden i ying acciden -in ol ed d i e s. (1992)/No quan ified pe iod o e ec s HIC: High-income coun y; MIC: Middle-income coun y; LIC: Low-income coun y. VTRA: Va iables ela ed di ec ly o RTA, in e ms o educ ions in he numbe o acciden s, casual ies, a ali ies o inju ed; VBD: Va iables ela ed o d i e beha io , in e ms o educ ions in isky o dange ous conduc , d i ing less unsa ely o a educ ion in law iola ions; VHCD: Va iables ela ed o heal h ca e da a, in e ms o educ ions in RTA- ela ed A&E admissions, su ge y o hospi aliza ions. Table 3 Selec ed s udies in he Eas e n-Medi e anean egion: coded da a. Sou ce: Au ho s. Coun y/ income le el Resea che / place P oxy o e ec s Me hodology Resul s Time se ies s udied/ ime-span o e ec s The Uni ed A ab Emi a es (HIC) Mehmood (2010)/(Al Ain) VBD Independen sample - es using SPSS. The DPS has s a is ically no significan impac on he speeding beha io o d i e s because o he lack o e ec i e a fic moni o ing sys em. (3 mon hs be o e/3 mon hs a e in oduc ion)/No quan ified pe iod o e ec s HIC: High-income coun ies; MIC: Middle-income coun ies; LIC: Low-income coun ies. VTRA: Va iables ela ed di ec ly o RTA, in e ms o educ ions in he numbe o acciden s, casual ies, a ali ies o inju ed; VBD: Va iables ela ed o d i e beha io , in e ms o educ ions in isky o dange ous conduc , d i ing less unsa ely o a educ ion in law iola ions; VHCD: Va iables ela ed o heal h ca e da a, in e ms o educ ions in RTA- ela ed A&E admissions, su ge y o hospi aliza ions. J.I. Cas illo-Manzano, M. Cas o-Nun ˜o / T anspo Policy 21 (2012) 191–201 195 possible he e ogenei y by conside ing ha he e ec s o expo- su e/in e en ion on he popula ion a e a ied and ha he e ec s analyzed by he included s udies a e only a andom sample o all possible e ec s. Weigh ing s udies wi h REM conside s no only hei own a iance (in a-s udy a iabili y), bu also he a iance ha migh exis be ween s udies (in e - s udy a iabili y): s y¼½Q2ðm1Þ cð3Þ whe e: Q¼De simonian and Lai d s a is ic. m¼no. o es ima ions conside ed. c¼es ima o calcula ed as c¼P m i¼1 w i ½P m i¼1 w 2 i =P m i¼1 w i whe eby he a iance and he s a is ical weigh o each i- h es ima ion would be, espec i ely n i ¼ s yþ i and w n i ¼1= n i In bo h cases (FEM and REM), he uppe /lowe limi s o he confidence in e als (CI) (in ou case, a 95%), a e calcula ed by he o mula (4): CI ¼exp Y71:96 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Ps a is ical weigh s p ! ð4Þ Any possible publica ion biases can be iden ified analy ically using he Begg (Begg and Mazumba , 1994) and Egge (Egge e al., 1997) es s, which es he null hypo hesis o he lack o said bias. Table 4 Selec ed s udies in he Eu opean egion: coded da a. Sou ce: Au ho s. Coun y/income le el Resea che / place P oxy o e ec s Me hodology Resul s Time se ies s udied/– ime–span o e ec s Czech Republic, The (HIC) Mon ag (2010)/ VTRA S anda d di e ences-in- di e ences me hodology. Fa ali ies we e abou 30% lowe du ing fi s h ee mon hs a e he law was in oduced. (2004–2008)/3 mon hs I eland (HIC) Bu le e al. (2006) VHCD S a is ical analysis using a wo- ac o ANOVA and Chi-squa ed es . In he fi s 6 mon hs, he e was a significan educ ion (48.4%) in RTA- ela ed spinal admissions. In he fi s yea , he e was a 25% educ ion in dea hs. (1998–2004)/6 mon hs Healy e al. (2004) VHCD S a is ical analysis wi h a Chi-squa ed es . The numbe o RTA- ela ed admissions ell o 17 compa ed o an a e age o 33 in he p eceding 4 yea s. (1998–2003)/6 mon hs Hussain e al. (2006) VHCD Re ospec i e coho s udy. Six y-one pe cen educ ion in ope a ions on inju ies caused by collisions. (2001–2003)/12 mon hs Saeed e al. (2010) VHCD S a is ical analysis was pe o med using he Chi- squa ed es . A e he o ense o no wea ing a sea bel was added o he DPS in 2003, he e was a 60% decline in he p opo ion o admissions o RTA- ela ed eye-inju ies. (2001–2007)/No quan ified pe iod o e ec s I aly (HIC) Benede ini and Nici a (2009) VTRA 3SLS eg ession and Poisson Reg essions. A s ong ‘announcemen e ec ’ wo yea s be o e he implemen a ion o he PPS. A dec ease o 72.87% in speeding icke s, o 17.80% in o al RTA and 26.42% in a al RTA. (2001–2008)/24 mon hs be o e in oduc ionVBD De Paola e al. (2010) VTRA Reg ession discon inui y design. A educ ion o abou 10% in RTA and o abou 25% and 15% in a fic a ali ies and inju ies, espec i ely. (2001–2005)/No quan ified pe iod o e ec s Fa chi e al. (2007)/(The Lazio Region) VHCD Poisson models. Twel e pe cen ewe RTA A&E isi s; hospi aliza ions dec eased by 16% and RTA- ela ed dea hs dec eased by 4%. (2001–2004)/6 mon hs Zambon e al. (2007)/ (Vene o Region) VBD Time-se ies analysis wi h ARIMA model. An inc ease in sea bel use o 51.8% among d i e s, o 42.3% among on passenge s and o 120.7% among ea passenge s. The numbe o RTA a ali ies and inju ies declined, wi h a educ ion o 18% and 19% espec i ely. (1999–2004)/18 mon hs VTRA Zambon e al. (2008)/ (Vene o Region) VBD Poisson models. An inc ease in sea bel use o 83% o d i e s and 76% o passenge s. (2003–2005)/15 mon hs Spain (HIC) Apa icio- Izquie do e al. (2011) VTRA Time-se ies analysis wi h ARIMA models. A mean educ ion in a ali ies o be ween 11.27% and 13.9%. (1995–2009)/E ec s main ained h oughou he s udy pe iod Cas illo- Manzano e al. (2010) VTRA Time-se ies analysis wi h unobse ed componen s model. An a e age educ ion o 12.6% in he numbe o dea hs in RTA. (1980–2007)/24 mon hs ( o dea hs) and 12 ( o inju ed) No oa e al. (2010) VTRA Quasi–Poisson eg ession models. The o e all numbe o people inju ed showed a educ ion o 5% and 4% among men and women, espec i ely, and an 11% and 12% educ ion o se iously inju ed people. (2000–2007)/18 mon hs Pulido e al. (2010) VTRA Time-se ies analysis h ough ARIMA models. A educ ion o 14.5% in o e all a ali ies. (2000–2007)/18 mon hs Uni ed Kingdom (HIC) Simpson e al. (2002) VBD Ques ionnai e su ey and empi ical analysis using Chi-squa ed es . The esul s o a new PPS o no ice d i e s led o a dec ease in he pe cen age o d i e s (5.8 o 4.5) o ending in hei 2nd yea o pos - es d i ing. (1992-94; 1997–1998)/ 24 mon hs HIC: High-income coun ies; MIC: Middle-income coun ies; LIC: Low-income coun ies. VTRA: Va iables ela ed di ec ly o RTA, in e ms o educ ions in he numbe o acciden s, casual ies, a ali ies o inju ed; VBD: Va iables ela ed o d i e beha io , in e ms o educ ions in isky o dange ous conduc , d i ing less unsa ely o a educ ion in law iola ions; VHCD: Va iables ela ed o heal h ca e da a, in e ms o educ ions in RTA- ela ed A&E admissions, su ge y o hospi aliza ions. J.I. Cas illo-Manzano, M. Cas o-Nun ˜o / T anspo Policy 21 (2012) 191–201196 This analysis mus be complemen ed wi h an in e p e a ion o so- called unnel plo s diag ams o a oid any alse posi i es o alse nega i es due o he loss o powe o de ec ion ha hese es s can be subjec o in small samples (S e ne e al., 2000). A unnel plo is a ool used o compa e he es ima ion in a g aph using some p ecision measu e, such as a s anda d e o unc ion, o example. I he g aph is symme ical, in an in e ed ‘‘V’’ shape, i is in e p e ed as p oo o he e p obably being no publica ion bias. I he g aph is asymme ic, i is in e p e ed as publica ion bias p obably exis ing. 3.3. Resul s Table 6 shows (in columns) he main esul s ob ained o he 4 me a-analyses o mula ed in all he scena ios (in ows), using a o al sample o m¼70 indi idual es ima ions o 26 sui able s udies. In o de o analyze he he e ogenei y o he sample he alues a e aken o he De simonian and Lai d Q (1986) s a is ics and he Takkouche e al. (1999) R I coe ficien (¼p opo ion o he o al a iance due o be ween-s udy a iance). The pooled odds a ios ob ained (OR) o each scena io wi h FEM and REM, espec i ely (wi h he confidence in e al limi s) a e se alongside he Begg and Egge es p- alues enabling he p esence o publica ion biases ha al e he final esul o be iden ified. All he Q s a is ics a e significan . This would, in p inciple, allow he homogenei y null hypo hesis o be ejec ed a 99% in all cases. Following Takkouche e al. (1999), gi en ha in he h ee analyses done o scena io I, R I is clea ly g ea e han 0.75, we unde s and ha he De simonian and Lai d Q s a is ic is su ficien ly powe ul o de ec he e ogenei y, o e coming he limi a ions ha his s a is ic p esen s o small samples (see Fleiss, 1993). The e o e, o scena io I, he OR es ima ed by REM is mo e eliable (which is why columns 7and8o Table 6 wi h he REM esul s ha e been highligh ed). This p esence o andom elemen s means i can be deduced ha he esul s o he in e en ion analyzed a e in eali y a ied, and ha he es ima ions o e ed by he a ious s udies included in ou me a- analysis a e only a andom sample o possible e ec s. The di e - ences om one coun y o ano he would be based on he coun ies’ own peculia ea u es, such as he ad e ising campaigns o he o he mo e o ce ul en o cemen measu es (fixed speed came as, lowe blood alcohol limi s, fines and hea ie legal sanc ions, poli- cing) ha accompany he applica ion o PS. In any case, in quan i a i e e ms no significan di e ences can be app ecia ed be ween he esul s ob ained wi h FEM and REM, which enables us o conclude ha ou es ima ions a e obus , as is Table 5 Selec ed s udies in he Wes e n-Pacific egion: coded da a. Sou ce: Au ho s. Coun y/ income le el Resea che /place P oxies o e ec s me hodology Resul s Time se ies s udied/– ime– span o e ec s China (MIC) Wong e al. (2008)/(Hong Kong) VBD Mul inomial logi model. An inc ease in penal y poin s p oduced a posi i e de e ence e ec in comba ing ed ligh iola ions (coe ficien ¼0.0959). (2005–2006)/12 mon hs Sze e al. (2011)/ (Hong Kong) VTRA Binomial eg ession model. The numbe o ed ligh o ense RTA and casual ies d opped by 23% and 29%, espec i ely. (2005–2007)/12 mon hs Aus alia (HIC) Diaman opoulou e al. (1997)/ (Vic o ia) VTRA Logis ic eg ession model. Adding a d i e ’s p io o enses (whe he as deme i poin le els o ca ego ies o o ense) in o his model p oduced he bes p edic i e abili y in iden i ying d i e s wi h u u e RTA. (1991–1994)/No quan ified pe iod o e ec s VBD Haque (1990)/ (Vic o ia) VBD Poisson model using maximum likelihood es ima ion me hods. The es ima ed mean in e -o ense ime in e al be ween 2 nd and 3 d o enses o d i e s was significan ly longe han he in e al be ween 1 s and 2 nd o enses. (1982–1985)/No quan ified pe iod o e ec s HIC: High-income coun ies; MIC: Middle-income coun ies; LIC: Low-income coun ies. VTRA: Va iables ela ed di ec ly o RTA, in e ms o educ ions in he numbe o acciden s, casual ies, a ali ies o inju ed; VBD: Va iables ela ed o d i e beha io , in e ms o educ ions in isky o dange ous conduc , d i ing less unsa ely o a educ ion in law iola ions; VHCD: Va iables ela ed o heal h ca e da a, in e ms o educ ions in RTA- ela ed A&E admissions, su ge y o hospi aliza ions. Table 6 Me a-analysis ou comes. Sou ce: Au ho s. Scena io Sample size Q s a is ic (p- alue) R Ia OR b (FEM)/ CI limi s 95% Resul d (FEM)/ CI limi s 95% OR c (REM)/ CI limi s 95% Resul d (REM)/ CI limi s 95% Begg’s es (p- alue) Egge ’s es (p- alue) I (VTRA) m¼24 0.0000 nnn 0.9944 2.9588 (2.9444; 2.9734) 19.2748 (18.993; 19.5583) 2.7570 (2.5516; 2.9789) 15.7525 (12.8276; 19.6662) 0.1574 0.3945 I (VBD) m¼18 0.0000 nnn 0.9998 3.4592 (3.4536; 3.4647) 31.7915 (31.6140; 31.9669) 3.3119 (2.9009; 3.7811) 27.4372 (18.1905; 43.8643) 0.2889 0.8765 I (VHCD) m¼8 0.0000 nnn 0.9992 4.0938 (4.0909; 4.0967) 59.9673 (44.1282; 61.6825) 3.9510 (3.7871; 4.1220) 51.9873 (59.7937; 60.1415) 0.7105 0.1318 II (LENGTH OF EFFECTS) m¼20 0.0002 nnn 0.6912 2.8088 (2.7305; 2.8893) 16.5900 (15.3406; 17.9807) 2.7663 (2.5602; 2.9840) 15.8997 (12.9384; 19.7667) 0.1119 0.2286 No es: Defini ion o a iables o Scena ios: VTRA: Va iables ela ed di ec ly o RTA, in e ms o educ ions in he numbe o acciden s, casual ies, a ali ies o inju ed; VBD: Va iables ela ed o d i e beha io , in e ms o educ ions in isky o dange ous conduc , d i ing less unsa ely o a educ ion in law iola ions; VHCD: Va iables ela ed o heal h ca e da a, in e ms o educ ions in RTA- ela ed A&E admissions, su ge y o hospi aliza ions; Leng h o e ec s: numbe o mon hs.Significance a nnn 1%, nn 5%, n 10%, espec i ely. a R I indica es he p opo ion o he o al a iance due o be ween-s udy a iance. b Odd Ra ios o fixed-e ec s model and andom-e ec s model exp essed in loga i hm. c Odd a ios o fixed-e ec s model and andom-e ec s model exp essed in loga i hm. d Resul s (exponen ial OR) exp essed in % o all a iables o Scena ios I and numbe o mon hs o Scena io II. J.I. Cas illo-Manzano, M. Cas o-Nun ˜o / T anspo Policy 21 (2012) 191–201 197 also shown by he sensi i i y analyses done. 8 In bo h models, he a iables ha expe ience a mo e posi i e e ec wi h he coming in o e ec o he PS a e he VHCD (wi h educ ions o o e 50% bo h in he numbe o hospi aliza ions and in A&E admissions), ollowed by he VBD (wi h a educ ion in he numbe o eckless beha io s o a ound 30% on a e age). Meanwhile, he weakes e ec seems o be on he VTRA (wi h educ ions in acciden s, a ali ies and inju ies o be ween 15 and 20%), i.e., he a iables mos di ec ly ela ed o oad a fic acciden s. Conside ing scena io II, a numbe o doub s could be aised abou which o he wo me hods o es ima ing he Leng h o E ec s is mo e app op ia e as, despi e he Q s a is ic ejec ing he homogenei y null hypo hesis a 99%, he R I coe ficien is sligh ly unde 0.75, indica ing ha i s de ec ion is less powe ul (Takkouche e al., 1999). No wi hs anding, his p oblem is i ele- an gi en how simila he es ima ions by bo h FEM and REM a e, wi h a di e ence o only 21 day be ween hem (16.6 mon hs compa ed o 15.9 mon hs e ec s du a ion). In any e en , he du a ion o he e ec s o PS is on a e age less han one and a hal yea s. When his esul is es ed agains Tables 2–5, i can be seen ha o he 26 assessmen s analyzed in 11 di e en coun ies, 20 de e mine he leng h o he e ec s o a PS, and o hese 9 gi e du a ions o o e 12 mon hs. E en in hese cases no e ec las ed a pe iod o o e 2 yea s, and he e was simply an in e im e ec compa ed o o he oad sa e y measu es wi h mo e s uc u al e ec s (see e.g., Apa icio Izquie do e al., 2011 o policing and Cas illo-Manzano e al., 2010 o manda o y sea bel use in Spain). Ano he in e es ing aspec is ha 7 o hese 9 e alua ions wi h e ec s o o e 12 mon hs a e o he same wo Eu opean coun ies, I aly and Spain, which s a ed ou wi h oad acciden a es ha we e highe han he mean o he Eu ozone (WHO, 2009). The sys em’s implemen a ion in hese coun ies was also accompanied by widesp ead dissemina ion o ad e ising cam- paigns in he ins i u ional media and in all p i a e news media, which shaped a a o able pe cep ion as a as public opinion is conce ned (No oa e al., 2010 o Spain; Benede ini and Nici a, 2009 o I aly). Specifically in he case o I aly, Benede ini and Nici a (2009) ound he e was an e ec e en be o e he PPS law came in o o ce, wi h he media in e es b inging abou a posi i e ‘‘announcemen e ec ’’, i.e., he e ec s o PPS in oduc ion we e an icipa ed e en be o e i had been e ec i ely adop ed as law. The emaining 11 s udies coincide in 12 mon hs being a qui e likely du a ion o he e ec s o he measu e and ha he e ec s would no emain s able h oughou his pe iod o a yea , bu would g adually dec ease. Specifically, a e a majo ini ial shock du ing he mon hs ha immedia ely ollowed PS coming in o o ce, he e ec s quickly wane wi h ime. The e is clea e idence ha his majo shock was no ed e en in coun ies whe e he e ec s immedia ely fizzled ou . In Canada, o example (see Redelmeie e al., 2003), al hough i was es ima ed ha he e ec s only las ed 2 mon hs, he e was a 35% all in he isk o a a al c ash and in he Czech Republic (see Mon ag, 2010), despi e he e ec s being limi ed o only 3 mon hs, a ali ies we e abou 30% lowe . The p- alues ob ained in he Egge and Begg Tes s in Table 6 show ha , in p inciple, ou es ima ions could be s able om he poin -o - iew o publica ion biases, as he lack o bias null hypo hesis canno be ejec ed in all he scena ios. Howe e , as hese de ec ion me hods a e known o no always be e y powe ul (e.g., S e ne e al., 2000), i is ecommended ha he es s a e complemen ed wi h an in e p e a ion o he unnel plo s diag ams (Palma-Pe ´ ez and Delgado-Rod ı ´guez, 2006) which a e included o all he analysis scena ios in Figs. 1–4. The pa ame e s anda d e o (SE) is ep esen ed on he x-axis, he pa ame e o each s udy (odds a io,OR) on he y-axis and each es ima ion by a poin . The esul is a unnel which na ows o he le as highe p ecision s udies (smalle SE, equi alen o a la ge sample size) a e on he le -hand side o he figu e and he e is less a iabili y be ween hem. A symme ical figu e ising om a ho izon al axis ha passes h ough he pa ame e ’s weigh ed alue would indica e he absence o his e o . As can be seen in Figs. 1–4, mo e o less asymme ical figu es a e ob ained o all scena ios which con adic he es s done in Table 6 and indica e he exis ence o publica ion biases. Howe e , cau ion should be used when in e p e ing his. In ac , acco ding o Tho n on and Lee (2000), symme y can be subjec i ely defined by he esea che , especially when he e a e small numbe s o poin s. He e ogenei y be ween he es ima ions can also in e e e when assessing o publica ion bias. No wi hs anding, ollowing Høye and El ik (2010) publica ion biases a e e y equen ly ound in me a-analysis o e alua e he e ec s o a oad sa e y measu e. These au ho s s a e ha he e a e ce ain p ejudices abou omi ing he publica ion o s a is ically nonsignifican s udies o hose wi h ad e se esul s, as i is no easonable o public opinion o expec a oad sa e y measu e ha has no e ec . Ano he possible sou ce o hese publica ion biases could be he s ic fil e applied o selec he s udy sample o he S anda d E o (SE) E ec Es ima ion Fig. 1. VTRA scena io. No e: E ec es ima ion in log scale. Sou ce: Au ho s. S anda d E o (SE) E ec Es ima ion Fig. 2. VBD scena io. No e: E ec es ima ion in log scale. Sou ce: Au ho s. 8 The esul s o he sensi i i y analysis a e a ailable om he au ho s upon eques . J.I. Cas illo-Manzano, M. Cas o-Nun ˜o / T anspo Policy 21 (2012) 191–201198 me a-analysis (see Sec ion 3.1). Ce ain s udies om se e al coun ies ha e been delibe a ely omi ed ei he because he me hodologies ha hey applied we e no e y igo ous (be o e and a e s a is ical compa isons), o because hey we e ‘suspi- cious’ o ha ing been ca ied ou by go e nmen al o ganiza ions and no being widely dissemina ed academically. 4. Conclusions and implica ions o oad sa e y Since hei pionee ing implemen a ion in he US in he mid 20 h cen u y, d i ing licenses based on poin s sys ems ha e apidly expanded o coun ies wi h high and medium income le els on all con inen s, especially in he Eu opean egion. As Nole ´n and O ¨s lin, (2008) and ETSC (2008) s a e, hei popula i y and social accep ance as one o he ai es co ec i e and punish- men measu es o in e nalizing he social cos s o oad a fic acciden s seem o be he key o hei wide- anging de elopmen o e he las decade. Despi e ecommenda ions om p es igious in e na ional o ganiza ions, e y li le is known in he academic li e a u e abou he e ec s o poin s sys ems on oad sa e y, because i is e y di ficul o isola e hei impac om he e ec s o o he concu en complemen a y ypes o en o cemen applied (SWOV, 2010) (policing, laws o ad e ising campaigns). In his ega d, s udies such as Cas illo-Manzano e al. (2010) poin o he mos con o e sial aspec o poin s sys ems linked o hei o e all e ec s in he sho - and long- e m. Ou s udy analyzes whe he poin s sys ems eally a e e ec i e o b inging down he a fic acciden a e o whe he his is jus a ashionable policy ha coun ies in close geog aphical p oximi y o wi h simila cul u al pa e ns imi a e. The e ec s o poin s sys ems and he du a ion o he e ec s ha e been es ima ed by applying a me a-analysis app oach o a sample o 70 es ima ions aken om 26 scien ific assessmen s om 11 coun ies. Al hough i migh esul in publica ion bias, a s ic sample selec ion c i e ion was op ed o and a numbe o suspicious and possibly less scien ifically igo ous s udies we e omi ed. The esul s o he sensi i i y analyses enable i o be obus ly s a ed ha all he a iables benefi om he implemen a ion o a poin s sys em. To be p ecise, educ ions o o e 50% in numbe s o oad a fic acciden - ela ed eme gency admissions, su ge y and hospi aliza ions a e eco ded. Reduc ions o a ound 30% in he numbe o law iola ions and isky and dange ous d i e beha- io s a e also es ima ed. Howe e , he mos significan e ec , (gi en ha he ou comes a e di ec ly linked o oad a fic acciden s), is he educ ion be ween 15 and 20% seen in he numbe o acciden s, a ali ies and inju ies. By con as , he high expec a ions ha poin s sys ems gene a e among public opinion and he oad sa e y au ho i ies wi h ega d o du a ion in ime ha e no been subsequen ly suppo ed by long- e m esul s as, in gene al e ms, he e ec s wea o in less han 18 mon hs. Only 7 o hese s udies s a e ha he e ec s ha e du a ions o o e 12 mon hs, bu also indica e ha hey las unde 2 yea s. Mos o hese se en s udies, inciden ally, e e o Spain and I aly, coun- ies, p ecisely, whe e he sys em has been accompanied by majo ad e ising campaigns, as shown by No oa e al., (2010) and Benede ini and Nici a, (2009). I seems ha his measu e has only a majo ini ial shock on a wide ange o oad sa e y a iables and indica o s, al hough he e ec quickly declines o e ime when he e a e no o he complemen a y en o cemen measu es. In sho , as Twisk and S acey (2007) s a e, many o he coun e - measu es ha a ec a licensing p ocess a e no e ec i e wi hou en o cemen ha inc eases he eeling ha you could be caugh , and complemen ed wi h media co e age ha educes he incen- i e o o ende s o pe sis in hei unsa e beha io s. Gi en ha he benefi s seem o be unning ou in e ms o oad sa e y, i is necessa y o compu e he es ima ed cos o imple- men a ion. To da e, li le has been p o ided ega ding quan ifica- ion o he han, acco ding o El ik and Vaa, 2004, ha hese cos s, a p io i, logically seem o be heo e ically lowe han o o he oad a fic acciden p e en ion me hods, such as oad imp o e- men s. Fo El ik. e al. (2009), he lack o a ull es ima ion o hese cos s would explain why o da e a mo e- han- equi ed Cos - Benefi Analysis has no been conduc ed o his measu e. Fo all hese easons, a u u e line o esea ch ha begs o be sugges ed is a ealis ic es ima ion o he cos s o his measu e. This es ima ion should ake in o accoun bo h he design o he in o ma ion campaigns ha suppo hem and changes in he coun y’s policing, legal and adminis a i e sys ems ha esul om hei applica ion. Wi h espec o he in o ma ion cam- paigns, hese ha e o be ex ended o co e he whole o he popula ion and also need o be in ensi e in o de o deal wi h all aspec s o he sys em, om he a ious iola ions ha a e en isaged, o he license wi hd awal p ocess and i s la e eco - e y. Wi h ega d o he policing and legal-adminis a i e sys ems being adap ed, o p e en injus ices he e should be ull coope a- ion be ween all he au ho i ies esponsible o oad sa e y wi h he aim o ensu ing ha he moni o ing, de ec ion and eco ding o all sanc ions is s anda dized whe e e hey ake place. I he measu e is implemen ed in a way ha is lacking in his espec i could lead ei he o he measu e being u ile, as would seem o ha e been he case in The A ab Emi a es (see Mehmood, 2010), o ha sys ema ic injus ices a e commi ed among S anda d E o (SE) E ec Es ima ion Fig. 4. VHCD scena io. No e: E ec es ima ion in log scale. Sou ce: Au ho s. S anda d E o (SE) E ec Es ima ion Fig. 3. VHCD scena io. No e: E ec es ima ion in log scale. Sou ce: Au ho s. J.I. Cas illo-Manzano, M. Cas o-Nun ˜o / T anspo Policy 21 (2012) 191–201 199