SURVEY ON FOREIGN TRAVELERS METHODOLOGY AND IMPLEMENTATION Rud Seljak 1, Nataša Jokć 2 1 Statstcal Offce of the Republc of Slovena, e-mal: rud.seljak@gov.s 2 Statstcal Offce of the Republc of Slovena, e-mal: natasa.jokc@gov.s Abstract In 2006 the Statstcal Offce of the Republc of Slovena started to carry out the Survey on Foregn Travelers, hch amed at obtanng data on the structure of foregn travelers leavng Slovena at road border crossngs. The man emphass as on foregn travelers ho dd not spend the nght n Slovena but came to Slovena just for a same-day vst (same-day vstors) or they crossed Slovena on ther ay to another country (transt travelers). The man target characterstcs ere the reasons for comng to Slovena and data on ther expendture. Snce the target populaton of the survey populaton as not knon n advance and there as no fxed samplng frame avalable, the sample selecton process and the estmaton procedure ere qute a challenge for our organzaton. The am of ths paper s to descrbe the methodologcal solutons used as ell as the concrete process of the executon of the survey. In the frst part of the paper the samplng procedure s descrbed, here the combnaton of stratfed random samplng and quota samplng as used. The second part s devoted to the descrpton of the estmaton process th the emphass on the usage of the external (auxlary) data for the mprovement of the fnal estmates. In the last part of the paper some concrete results and ther coherence th the results of other surveys ll be presented. Keyords: flo samplng, estmaton methodology, travelers statstcs 1. Introducton Data on toursm statstcs are hghly relevant not only for toursm polcy makng but also for other related felds of statstcs such as the Balance of Payments. The mportance of toursm n natonal economy s evaluated by Toursm Satellte Accounts (TSA). Besdes toursts 1, same-day vstors and transt travelers have an mportant role n toursm ndustry. In order to comple the data sources for Travel Balance of Payments on behalf of the Bank of Slovena, the Statstcal Offce of the Republc of Slovena (SORS) carred out the Survey on Foregn Travelers at Road Border Crossngs. The man purpose of ths survey as to obtan the data on the structure of foregn travelers leavng Slovena at road border crossngs n terms of ther reasons for comng to Slovena. The man emphass as on foregn travelers ho dd not spend the nght n Slovena but came to Slovena just for a same-day vst (same-day vstors) or they crossed Slovena on ther ay to another country (transt travelers). Besdes the structure of foregn travelers, the purpose of the survey as also to obtan the data on expendture of foregn same-day vstors and transt travelers. The survey as conducted for the frst tme n summer 2006 only at the border th Croata. The results of ths survey ere not publshed but ere only used for calculatng the Balance of Payments. In Aprl 1 Toursts are persons travelng to a place other than that of ther usual envronment, stayng n a hotel or some other tourst accommodaton n the place/country vsted for at least one nght (but not more than 365 nghts consecutvely) for lesure, busness or other reasons, but not to make proft n the place vsted (they receve no remuneraton) (Applyng Methodologcal Gudelnes n Basc Toursm and Travel Statstcs, 1996).
2007 and n the perod beteen 15 th July and 15 th August 2007 the regular survey as conducted at all four borders (Italy, Austra, Croata and Hungary). The most challengng problem hch had to be dealt th n the phase of survey plannng as the queston ho to desgn an effcent and yet easly feasble samplng procedure. Whle n most of our samplng surveys the procedure follos the classcal procedure of samplng frame constructon and then the sample selecton by one of the ell knon samplng desgns, the stuaton here as qute dfferent. Snce the travelers crossng the border crossngs are the classcal example of flos, e had to deal th the problem of the flo-samplng and the drect consequence of that as the fact that there as no fxed samplng frame avalable. We could just use some auxlary data to predct the sze of the flo of the travelers n the dfferent days at the dfferent border crossngs and n fact these auxlary data played the crucal role n the samplng procedure. The hole samplng desgn as at the end the combnaton of mult-stage PPS samplng and quota samplng. Snce the target populaton as not represented through the fxed samplng frame, also the nferental procedure (mostly represented by the eghts calculaton) as qute challengng. Although the bass approach as stll n the drecton of the ell-knon Horvtz-Thompson estmator, some addtonal smplfcatons ere needed. In the frst part of the paper the samplng desgn and the eghtng procedure ll be descrbed. In the second part some man results ll be presented and at the end some conclusons ll be ponted out. 2. Samplng Desgn The am of the samplng procedure s to get the representatve sample of the foregn travelers ho have crossed one of the Slovenan border crossngs thn the reference month. Hence the samplng procedure has to deal th to dmensons: spatal (border crossngs) and temporal (days, hours, etc.) dmenson of the target populaton. We present the stuaton n the follong chart here the abbrevaton BC s used to desgnate the dfferent border crossngs and t represents the tme dmenson, assumng that e have n border crossngs and m tme-unts n the target populaton. Fgure 1: Vsual presentaton of the target populaton BC 1 (t 1 ) BC 1 (t 2 ) BC 1 (t m ) t BC n (t 1 ) BC n (t 2 ) BC n (t m ) The arros n both drectons represent the ncomng and outgong flo of the travelers. In fact, only n the frst executon of the survey both drectons ere sampled. Snce the analyses of the results of
ths frst executon shoed no sgnfcant dfference accordng to the dfferent drecton, n all the subsequent executons only the outgong flo as surveyed. As already stated, our fnal target unt as the foregn traveler crossng one of our road border crossngs. Snce t s qute understandable that the lst of vehcles or travelers as not avalable at the tme of settng up the samplng procedure, the classcal random samplng could not be used. Frstly, e had to defne several stages of samplng, collect the auxlary nformaton for each of the stages and on the bass of the avalable nformaton determne the fnal samplng desgn. The most mportant auxlary data used n the samplng procedure ere the data on the number of vehcles crossng the border (n both drectons) for every hour and every day of the prevous year. These data ere provded by the Slovenan Roads Agency (SRA) and obtaned by the use of the automatc counters. Snce t s qute obvous that the number of vehcles crossng the border s the subject of sgnfcant peaks through eekends and holdays, the counts from the prevous years had to be correctly mapped to the current-year day. Hence, for each prevous-year day d the count ( d +366) as used as the estmated flo sze for the current year. These data ere n the samplng stage used to determne the sze of the border crossng as ell as for all the PPS procedures. The samplng procedure could roughly be dvded nto to separate parts, here for the frst part classcal PPS random samplng and for the second part quota samplng as used. Snce the lst of the real observatonal unts as not avalable at the tme of samplng, e had to deal th the ndrect unts. The populaton of travelers could be conceptually consdered as a to-dmensonal flo U ( x, t), here x represents the spatal and t the temporal dmenson of the flo. If e ant to get at least approxmately representatve sample, e have to sample n both dmensons. The only possble samplng unt for the spatal dmenson s clearly the border crossng; therefore m 2 border crossngs for each of the drectons ere selected at the frst stage. The populaton of all border crossngs as frstly stratfed accordng to the border state 3 and accordng to the sze. Snce the border crossngs are qute heterogeneous n the sense of the number of travelers, e took the take all approach, usually used n busness surveys. The crossngs that ere put n the stratum of largest crossngs ere selected th certanty. From the smaller crossngs strata the PPS random sample as selected. For the temporal dmenson, the to stage samplng as used. At the frst stage the samplng unt as the calendar day and e agan used the PPS samplng to select the allocated number of days for each of the selected border crossngs. As t as decded pror to the samplng selecton, each selected day one nterveer should survey at the border crossng for eght hours. Therefore the second stage unt as the hour of the begnnng of nterveers ork and for each selected day the startng hour as randomly selected by usng the PPS samplng agan. So, the hole frst part of the samplng procedure could be consdered as a three stage samplng, here for each of the selected border crossngs the survey days ere selected and then as the thrd stage for each of the selected days the startng hour as selected. In the second part of the samplng procedure the vehcles and then the travelers had to be selected. Snce t as qute dffcult to perform a correct random procedure n the stuaton lke ths, e decded to go for a quota samplng at ths stage. Each nterveer as thus nstructed to get at least 80 completed questonnares for each of the 8-hours shfts. Later n the stage of eghts constructon, the quota samplng at the last stage as consdered as the smple random samplng here all the vehcles had equal chance of beng selected. We also present the system graphcally. 2 The number of selected border crossngs slghtly dffers n dfferent executons of the survey but t as generally beteen 20 and 25. 3 Slovena has four border states: Austra, Croata, Hungary and Italy.
Fgure 2: Vsual presentaton of the samplng procedure 3. Weghtng Snce all the data ere collected on the bass of (mostly) randomly selected sample, all the data should be properly eghted n the estmaton procedure. The eghts calculaton should as strctly as possble follo the selecton procedure n order to assure that the eghts and consequently the fnal results should properly reflect the effectve sample as ell as some exogenous nformaton on the target populaton. We descrbe the concrete procedure through the follong steps: Weght for the selected day. In the frst step of the procedure the samplng eght for each of the selected days as calculated. Follong the selecton procedure ths eght should be calculated as the product of the to Horvtz-Thompson eghts; the eght reflectng the selecton probablty of the border crossng and the eght reflectng the selecton probablty of the selected day for the partcular border crossng. Hence, for each day d for the selected d border crossng BC, the eght s calculated as: = 1 π ( BC )) (1 π ( d )), here π (.) j ( j denotes the selecton probablty. Weght for the selected traveler. Whle n the frst stage of the eghts calculaton only the nformaton derved drectly from the samplng procedure as used, for the second stage, hch should reflect the quota samplng, some addtonal assumptons ere needed. For the sake of smplcty, e assumed that the quota sample as n fact a smple random sample of all foregn travelers crossng a partcular border on a partcular day. So, the eght at ths stage should be calculated as the rato beteen the number of all the foregn travelers crossng the border and the selected foregn travelers. Snce the number of selected travelers as of course knon from the survey, the populaton count should someho be estmated. Agan here the automatc counters data ere used but n ths case already the data from the current year ere avalable, hence the counts for the exact date of survey executon could be used. We then combned these data together th the estmatons of the share of the foregn vehcles for each surveyed border crossng hch e obtaned from the Cross Border Survey. Snce these estmatons gave us only the number of foregn vehcles, e stll needed to expand these counts to the number of foregn travelers. For ths purpose e estmated the average number of travelers n dfferent types of vehcles (cars, buses, vans, etc.) drectly from the survey data and then used these averages to get the anted estmates. The eght for the partcular traveler j
on the partcular day j at the partcular border crossng as hence calculated by the formula: V T k= j = 1 N kj T n j k f k here V s the number of dfferent types of vehcles, N kj number of counted vehcles of partcular type, T k average number of travelers per vehcle, fk estmated share of foregn vehcles and n number foregn travelers nterveed. The fnal samplng eght for the j selected traveler nterveed on the day j at the border crossng as then calculated as s j =. d j T j Calbraton. Snce the frst step of the samplng eghts calculaton as entrely based on the populaton counts from the prevous year, also the fnal samplng eght partly reflects the populaton from the prevous year. Therefore n the last step the eghts ere calbrated to the current year populaton counts. The calbraton as performed on the bass of the estmated populaton counts of the foregn travelers for each eek of the surveyed month. Usng the denotatons from above, for each eek these estmates ere obtaned by usng the formula 7 BC V N = N T f. If n s the number of the foregn travelers nterveed durng j= 1 = 1 k= 1 kj k k the partcular eek, the fnal eght s then calculated as: N s = n. s =1 The above gven descrpton descrbed the procedure for the calculaton of the eghts referrng to the selected traveler. In fact, n the concrete procedure besdes the eghts for the travelers also the eghts for the selected vehcles ere calculated. The calculaton procedure for these eghts as mostly n accordance th the procedure for the travelers eghts, the only dfference as that here the usage of the average number of travelers per vehcle as omtted. The eghts, calculated as descrbed, ere used as the basc tool, hch enabled the correct nferental procedure for the calculaton of the statstcal estmates. Although the specfc stuaton of the flosamplng demanded some smplfcatons hch caused a slght departure from the strct theoretcal concepts, all the analyses proved that for the most of the estmated fgures a suffcent degree of relablty has been assured. In order to separate the estmates th loer degrees of relablty, e also estmated standard errors for each estmated statstcs and then n accordance th the general polcy of SORS (see Seljak(2008)), some estmates th the largest standard errors ere not dssemnated and some estmates ere marked as the estmates of loer precson. 4. Results The questonnare as dvded nto to parts: the frst part related to some man characterstcs (country of resdence, type of vehcle, enterng Slovena at hch border, man reason for comng to Slovena). If the man reason for comng to Slovena as a same-day vst or transt, the foregn travelers ere asked about ther expendtures durng the vst n Slovena. The data on expendture referred to the hole famly or travelng party th hom the respondent traveled and shared costs.
As regards country of resdence, most of the foregn travelers came from countres that border th Slovena (Italy, Austra, Hungary and Croata). In Aprl 2007 three quarters of foregn travelers ere from those countres. In the summer months the natonal structure of foregn travelers as more dverse (55% of travelers ere from neghborng countres, 17% from Germany and 28% from other countres). The share of foregn travelers accordng to the man reason for comng to Slovena dffered n the sprng months from those n the summer months. Whle n Aprl less than half of foregn travelers ere n transt, n summer months the share of transt travelers as about to thrds. In Aprl one thrd (33%) of foregn travelers came to Slovena for a same-day vst, hle n July and August 16% of foregn travelers came for that reason. The share of foregn travelers ho stated other reasons for comng to Slovena (stay of several days, ork, school, daly or ordnary busness) as n Aprl and n summer months approxmately the same (about 18%). Fgure 3: Foregn travelers by man reason for comng to Slovena, Aprl, July-August 2007 TRANSIT 49% OTHER 18% Aprl 2007 SAME-DAY VISIT 33% same-day trp 48% cultural ev ent 4% shoppng 29% other 19% TRANSIT 65% July-August 2007 OTHER 19% SAME-DAY VISIT 16% same-day trp 49% cultural ev ent 7% shoppng 32% other 12% Source: SORS, Survey on Foregn Travelers at Road Border Crossngs, Slovena, Aprl, July-August 2007 We asked same-day vstors about the man reason for ther same-day vst to Slovena. Less than half of them came on a same-day trp (e.g. for entertanment, sports, sght-seeng, alkng, lunch, etc.), less than one thrd (29% n Aprl and 32% n July and August) came for shoppng, hle less than one tenth n July and August stated other reasons (above all vsts to relatves and frends, busness meetngs). Most same-day vstors ere from Italy, folloed by same-day vstors from Croata and Austra. Same-day vstors (spendng about 35 euros per person n Aprl and about 43 euros per person n July and August) spent on average about four tmes more than transt travelers (about 9 euros per person per day n Aprl and about 11 euros per person per day n July and August). As regards the man reason for same-day vst, most money as spent by same-day vstors that came to Slovena for shoppng. 4.1 Coherence of the results th the other survey Many European countres are aare of the mportance of same-day vstors and they carry out surveys to collect nformaton on them (some countres for nbound, some for outbound same-day vsts and some for both). Even though there are some basc gudelnes about the defnton of same-day vsts, countres are usually usng slghtly dfferent defntons of same-day vstors consderng ther specfc stuaton. Countres are collectng data on same-day vstors for dfferent reference perods. Consderng ths and the fact that country breakdon n some surveys s not suffcent, mrror statstcs s hard to use. Countres usually publsh same-day vsts of domestc populaton dvded nto domestc same-day vsts and same-day vsts abroad and do not break don same-day vsts abroad accordng to country of destnaton. In many cases breakng don data on same-day vsts abroad accordng to country vsted makes data for certan countres less relable.
Due to specfc characterstcs of transt traffc, a comparson th other countres s not possble. Wth the ntenton of fndng out the coherence of the data the results of the Survey on Foregn Travelers at Road Border Crossngs ere compared th the results of the Cross Border Survey hch monthly montors cross-border traffc of vehcles, passengers and goods by road. Enterng Slovena n the Schengen area made t possble to carry out the survey only at the border th Croata. That s hy e compared data beteen these to surveys only for the border th Croata. Ths comparson shoed certan dstncton consderng the number of foregn travelers crossng the border n Aprl. Accordng to the Survey on Foregn Travelers, the number of foregn travelers leavng Slovena at road border crossngs as n Aprl 2007 13% loer than the number of foregn travelers accordng to data of the Cross Border Survey. In July-August 2007 the dstncton as neglgble (less than 1%). We assume that ths dfference s on one hand the result of the use of a slghtly dfferent methodology for collectng the data. The data on road cross-border traffc of vehcles and passengers are provded by the border polce ho at some border crossngs estmate the number of passengers usng nformaton on the number of vehcles crossng the border and certan coeffcent about average occupancy of the vehcle. In the Survey on Foregn Travelers at Road Border Crossngs nterveers had to count the number of travelers n the vehcles that ere ncluded n the survey. Comparng average occupancy rate of the vehcles shos that the Survey on Foregn Travelers ascertans certan dstncton beteen Aprl and summer months, hle accordng to the Cross Border Survey the average occupancy rate as almost the same n both observed perods of the year. We assume that data on average occupancy rate of the Cross Border Survey ere a bt overestmated n the month outsde the hgh tourst season (Fgure 4). Fgure 4: Average occupancy rate of prvate vehcles 4 at the border th Croata, Aprl, July-August 2007 (data for the Cross Border Survey are unoffcal and are only for the purpose of comparson n ths paper). 3 2,5 2 1,5 1 0,5 0 Aprl July -August Surv ey on Foregn Trav elers Cross Border Surv ey Source: SORS, Survey on Foregn Travelers at Road Border Crossngs, Slovena, Aprl, July-August 2007 SORS, Cross Border Survey, Aprl, July, August 2007 On the other hand, data from the Survey on Foregn Travelers at Road Border Crossngs are based on a sample, hch means that certan sample errors alays occur. 5. Conclusons The Survey on Foregn Travelers, hch has been carred out by SORS snce 2006, meant a challengng task at the tme of plannng and desgnng as ell as later n the executon phase. Snce 4 For ths comparson as prvate vehcles ere consdered prvate cars, vans, campervans.
several characterstcs of the survey (especally the target populaton of the flo of travelers) ere qute dfferent from other regular surveys conducted at SORS, some ne approaches ere needed. In the frst part of the paper e descrbed the samplng procedure, hch could roughly be dvded nto to parts. In the frst part e used the prevous year counts to predct the present year flo sze and then on the bass of ths predcton used PPS samplng desgn to select border crossngs, days and startng hours for the feld nterveers. In the second part of the procedure the nterveers ere nstructed to obtan a certan quota of completed questonnares by nterveng the foregn travelers n dfferent types of vehcles. The collected data ere then statstcally processed; the grossng-up eghts ere calculated and th the use of these eghts the populaton estmates ere produced and fnally dssemnated. To basc precondtons ere needed to assure estmates th a suffcent degree of accuracy. The frst s the assumpton that the flo-szes do not dffer much accordng to dfferent years. The second one s the assumpton that the nterveers strctly follo the nstructons, especally the nstructons on the allocated survey date and the startng hour on the selected day. Any departure from the planned samplng scheme requres use of the adjustng procedures n the eghtng procedure and can cause a loer degree of accuracy. One problem for conductng the border survey n the future s the fact that Slovena entered the Schengen area th the other ne Member States of the European Unon n the second half of December 2007. In our opnon the most mportant data as the output from conductng the survey at all natonal borders n 2007 are the estmates of the transt traffc flo and data on the number of transt travelers (e estmate that most of the transt traffc s orented to the border th Croata). Even though conductng the survey only at the border th Croata ll be suffcent for certan tme, e are plannng to carry out surveys near other natonal borders th the dfference that nterveers ll move from border control ponts to the nearest gas statons, parkng areas or supermarkets. Data on same-day vstors ll also have to be completed th a survey at mportant tourst attractons around the country. References [1] Ksh Lesley (1965) Survey samplng: John Wley & sons [2] Sarndal Carl-Erk (1992) Model Asssted Survey Samplng: Sprnger-Verlag [3] Seljak Rud (2008) Standard Errors Presentaton and Dssemnaton at the Statstcal Offce of the Republc of Slovena, European Conference on Qualty n Offcal Statstcs, Rome [4] Applyng the Eurostat Methodologcal Gudelnes n Basc Toursm and Travel Statstcs: A Practcal Manual (revsed verson): March 1996