MPLEMENTATION OF A NATIONAL OBSERVATORY FOR MONITORING TECHNO-ECONOMIC DATA OF THE ITALIAN FLEET AND THE EVALUATION OF SOCIO-ECONOMIC PARAMETERS 1 1. Data collection and estimates of economic parameters concerning te Italian fising fleet is produced by IREPA troug a National Observatory, wic dates back to te early 80 s. Oter sources of statistical information also exist; in particular te National Institute of Statistics (ISTAT) is involved in te production of estimates concerning some of te main figures of te Italian fisery sector. 2. In 2000 Italy started a process of rationalisation and armonisation of te existing surveys on te fisery sector. Tis process was aimed at solving one of te most important problems debated in te national and international working groups tat is te availability of different and disagreed statistical sources for te same penomenon. Te practical outcome of tis process was te definition of a sample survey on te catces and te relative values wose objective is to satisfy te EU legislation and, more in general, te national and international information needs. Te metodology of te new survey as been developed by IREPA Onlus, in collaboration wit ISTAT (te national statistical institute). In future, te Ministry for Agricultural and Forestry Policy will be responsible for te surveys and for te publication of statistics. In two years, tis new approac sall replace te surveys currently carried out by ISTAT and by IREPA and it sall represent te only official statistical source on te sector. Te proposed sample survey is described in te following pages. 1. AIMS OF THE NEW MONITORING SYSTEM 3. Te statistical survey s aim is to gater information on te most significant parameters of te fisery sector. 4. To tis aim te existing observatory will be improved and it will consist of tree main modules: Module of evaluation of fising effort and activity. Module of evaluation of landings and prices by group of species. Module of evaluation of economic and social data. 5. Te survey is based on a unique sample. More ten 750 vessels are monitored eac week and elementary data are later expanded to te universe (te wole Italian fleet) using statistical sampling procedures. 1 Te observatory as been implemented by IREPA Onlus Istituto Ricerce Economice Pesca e Acquacoltura, and it is partially funded by te FIFG programme, under te tecnical assistance measure. 1
6. It is wort underlining tat te researc programme on systematic monitoring of fisery indicators in Italy as targeted and still targets an evaluation of economic and management features of fiseries. It does not aim to estimate and assess biological resources. DESCRIPTION OF THE SURVEY Assessment of te sampling universe and of te list 7. Data collection concerning te fisery sector in Italy is very complex due to te ig number of species caugt, te spreading of te fleet along te 8000 Km of coastline and te vast number of landing points available (estimated at around 800). 8. Te National Fleet Register (Arcivio Licenze di Pesca ALP-) constitutes te list from wic te sampling units are extracted were all Italian fising vessels are included. Te Fleet Register is eld at te General Directorate for Fiseries and Aquaculture of te Ministry of Agricultural and Forestry Policies (Direzione Generale Pesca del Ministero delle Politice Agricole e Forestali). 9. Terefore, te sampling basis used for te survey (2001) is te Fleet Register updated to June 2000. In tis arcive te total number of fising vessels in Italy is sligtly iger tan 18 tousand units. Twenty-seven vessels fis beyond te straits (te so-called oceanic fleet). Considering te specific activity of te latter vessels, tey ave been excluded from te sampling basis and data referring to tis segment are gatered on te basis of taxable property, by means of agreements wit sip owners and teir category association (Federpesca). Tuna fising vessels (associated in te Associazione Produttori Tonnieri Salerno) are also excluded from te sampling basis and data concerning teir activity are provided directly by te Association for its member vessels. Te questionnaire and te coice of te data collectors 10. Sample data are recorded by means of tree specific questionnaires: An annual questionnaire to record tecnical, dimensional and vessel management information on te sample units and relevant socio-economic aspects (number of sipowners, teir ages, teir property quotas and relationsips between tem). A quarterly questionnaire to record data on fixed and variable costs, and on social aspects of property and crew. A weekly questionnaire to record information reporting activity suc as fising time and area, average number of crewmembers, gears used, quantities, prices and revenues as per species or group of species and trade cannel for sales. 11. Te questionnaire for te survey as been designed so tat te questions sequence can be defined as "funnel-saped". Te first part of te questionnaire concerns general information suc as te name of te boat, te gears employed, te days of activity at sea, te days of bad weater, non-fising days for rest, te total number of ours, te average crew and te distance of te fising area from te coast. Te second part is meant to gater te survey s target information suc as te species caugt (quantity, quality, average prices, destination). 12. Te questions are structured according of te caracteristics of te penomenon and te degree of knowledge. In oter words, tere is no need to coose between open-response questions and fixed-response ones. In particular, an exaustive list of te species for wic quantity and prices are 2
required as been prepared, and te data collectors duty is to report te individual species caugt and teir prices. 13. Oter important aspects of te questionnaire s design, suc as te use of te language, te formulation of te questions, te correct reporting of information, are andled directly by te data collectors, tat terefore represent a filter between te people interviewed and te data processing unit. It is also to be noted tat te input of data for te single vessel is fully computerised; te software, specifically designed for te survey s objectives, is logically structured and also includes crossceck programmes to avoid partial or inconsistent filling of te questionnaire. 14. In brief, te most important annual, montly and weekly informations recorded are te following: Annual information Name Gross registered tonnage (GRT) Maritime district were te boat as been registered, (coastal area/sector) Gross tonnage (GT) based on London Convention (Reg. EC 2930/86) First year of service (terefore, age) Horsepower (kw) Autorised fising gears Engine make, location and type of propeller Maritime district from were te sip departed for fising Communication engine Maritime district were te product is landed Navigation engine Type of association and year of its creation Number of sip-owners, teir ages, teir property quotas and relationsips between tem Type of association and year of its creation Fis location engine Conservation equipment Employment contract used Lengt overall and lengt between perpendiculars Quarterly information Name Fis transport cost Mont Oter running cost Maritime district were te boat as been Labour sare, wages and social insurance registered (coastal area/sector) Fuel (total and unit value) Cost of nets Cost of bait Cordage and ropes Food Boxes and ice Commercialisation costs Oter running costs Ordinary maintenance Extraordinary vessel maintenance Extraordinary ull maintenance Extraordinary engine maintenance Vessel insurance Tax and oter fiscal costs Bank carges Oter vessel costs 3
Name Weekly information Non fising days for rest, repair and oter Week Maritime district were te boat as been registered Hulls Average time (in ours) for eac single trip Engine used Average crew Fising days Total ours at sea (navigation and fising) Minimum and maximum fising area s distance perpendicular to coast line Maritime district from were te sip departs Maritime district were te product is sold For eac single species or group of species landed: quantity, prices, income and commercial cannel (wolesaler, fis market, retail dealer, oters). Non fising days for bad weater 15. Data collectors belong to te productive or management fisery sectors. Tis is certainly an innovative element for official surveys. As matter of fact, data collectors are usually external to te penomenon tat is being examined and, moreover, tey are often part of some public structure, in order to avoid possible influences due to personal interests. On te basis of te experience acquired in tis field, it as been demonstrated tat it is essential to ave data collectors belonging to te fisery productive cain in order to obtain correct and timely data. Obviously, periodic inspections in te various areas are carried out in order to ceck te data collectors work. 4
Monitoring system - survey points SAVONA GENOVA IMPERIA PORTO TORRES VIAREGGIO LIVORNO PORTOFERRA IO CIVITAVECCHIA ROMA GAETA NAPOLI VENEZIA CHIOGGIA RAVENNA RIMINI PESARO ANCONA S.BENEDETTO d.t. SALERNO PESCARA MANFREDONIA MOLFETTA BARI VIBO VALENTIA CAGLIARI CROTONE PA LERMO TRAPANI MA RSA LA MA ZARA SCIACCA MILAZZO MESSINA CATANIA SIRACUSA Sampling design: Single Stage Stratified Sampling 16. Te sampling design is based on a single stage sample stratified over two variables. Stratification is carried out in order to create strata as omogeneous as possible, using caracteristics correlated to te target variables. 17. Te maritime regions of te Italian coast represent te first stratification variable from an administrative point of view. Te sampling design considers only 13 of tese 15 maritime regions, since tere are no enrolment offices in Basilicata and, due to te small size of Molise s fleet (0.3% of te total number of Italian fising vessels), te latter as been aggregated to Abruzzo. 18. Te second stratification variable is te fising system used by te vessel; te fising fleet is terefore divided in groups and eac fising vessel belongs to one of tese groups according to te fising system it uses. For te present surveys te following systems are considered: trawlers, dredges, drift-nets, long-line, multipurpose and small scale fisery. Te identification of te fising system follows te fising systems actually present in te Italian fleet, wile taking into account criteria of consistency wit te segmentation considered under te MAPG IV (Multi Annual Guidelines Programme). 19. Furtermore, for some systems (trawler and multipurpose) and for some regions (Abruzzo, Marce, Puglia, Sicilia and Veneto) anoter stratification is carried out on te basis of te dimensional parameter GT (Gross Tonnage). Tis in order to obtain more omogeneous study domains and to take into account te differences between te coastal or sort range trawler and te ig sea trawler. 5
20. Moreover, for dredges registered in Veneto, Emilia Romagna and Marce regions, anoter stratification as taken place, based on te marine compartments to wic tey belong. Tis decision as been taken in order to consider te peculiar situation of tis fising system tat, due to te actual national regulation wic allows for a self management approac, at te compartment level by te Bivalve Molluscs Management Consortiums (Consorzi di Gestione dei Mollusci bivalvi). 21. Finally, tose vessels autorised to fis tuna among oter species represent an additional specific stratum. Tis stratum is identified in reference to te 7/2/2000 ministerial decree regarding te "determination of individual fising quotas for blue-fin tuna for te year 2000", according to wic a complete list of te units operating on tis target species is reported. Te vessels of te Associazione dei Produttori Tonnieri di Salerno do not belong to tis segment and, as said before, tey are not sampled and belong to a specific stratum, since tey fis exclusively tuna. 22. Te final number of strata or domains from wic te overall sample is extracted amounts to 70. To te latter, te two strata tat are out of te sample (Oceanic fleet and Associazione dei Produttori Tonnieri di Salerno) need to be added. Sample size and allocation across strata 23. Te size of te sample is assessed on te basis of te evaluation of te sampling error. 24. In particular, tis implies te specification of te estimates reliability, wic consists in setting te value of te average square error; furtermore, since correct or approximately correct estimators are used, te values of te estimates variances are determined. 25. In te case of single stage stratified sampling, and in te ypotesis of extracting te sampling units wit equal probability and witout re-pooling, te sampling variance of te estimate RIWKHWRWDO< is described by te following expression: ^ V ( Y ) H 2 2 N n S N, 1 N n given tese definitions: H represents te number of strata in te population ^ Y represents te estimated total of Y for te population N represents te total number of sampling units in te t stratum n represents te total number of sampling units measured in te t stratum 2 S represents te variance of Y for te t stratum 26. Terefore, for a given population, variance varies bot as a function of te overall size of te sample n and, for a set value of n, as a function of te numerosities n 1,, n,, n H, wit te imposed constraint tat teir sum must be equal to n. 27. Among te various ways of determining te sampling sizes of te H strata, te Neyman criterion as been used instead of te proportional one. Te Neyman metod is based on te criterion by wic a different percentage of elements is drawn from eac stratum in order to minimise te value of V( 6
28. Neverteless, te Neyman metod can be applied only in te case of a single target variable, oterwise we would obtain a different sample size for eac variable considered. Since te present survey is multivariate, tat is, te variables investigated are more tan one, to calculate te sample size we use te Betel metod, wic is te application of Neyman s metod to te multivariate case. Te approac used by tis metod is to transform te analysis into a linear programming model tat allows to identify te sample size and te allocation across strata, minimising te variances of all variables simultaneously (see also Betel, 1989). 29. Te optimal allocation across strata for multi-scope studies as been solved by Betel using te Kun-Tucker teorem and ten deriving te expressions for te optimal allocation in terms of te LaGrange multipliers. 30. In order to apply tis metod, a pre-estimate of te S 2 variances is required; in oter words, te variances of te target variables of te survey must be known. For tis purpose, te results of a sampling survey conducted in 1999 on te production data of more tan 400 vessels ave been used. It was decided tat te variables to be considered to calculate te sampling size, must be tose for wic te variance is igest; catces by species and by region were cosen. So, to apply te Betel model montly estimates of te total catces by species and by region must be known. 31. Te Betel metod as been applied to te data available for 1999, wit a procedure specifically implemented on SAS basis. 32. In a first pase, since te sample size is inversely proportional to te error level, tree different levels of sampling, wit tree different levels of maximum acceptable error, ave been identified. However, te final coice as been te lowest level of te maximum acceptable error (average sampling error of 5%, on a confidence interval of 95%), wit a total sample size of 440 units and a degree of coverage of 2.4%. 33. Te final size as been obtained by applying te Betel procedure wit a constraint of minimum size per stratum of 4 units (wit te exception of te stratum Campania-mecanised dredge; te numerosity considered for tis stratum is te one tat derived from te application of Betel s algoritm). 34. Te sample units ave been extracted by applying te PPS metod (proportional to te size metod). Eac unit as a different probability to be sampled and tis probability is proportional to te following measure: LFTi P i LFT were: i = a generic vessel = stratum LFT= overall lengt. Among te different metods to extract a sample, te Hanurav algoritm was cosen. 7
Expansion factors 2 35. In brief, to pass from te sampling data to te overall estimates, te weigt applied to eac elementary data is te following: 1 1 LFT k i S LFT i i nlfti n LFT were: S i : te probability of te unit i to be selected n : sample size for te stratum LFT: cumulative vessel lengt. 36. In te case of non-responses, te initial weigts k i are adjusted on te basis of data referring to te responses (r ) and te non-responses (s ) of te sample (n ). Te metod consists in multiplying te initial weigts (k i ) by a factor (d ) equivalent to: r s d r 37. Te resulting weigts (v i ), tat are called base weigts, because tey are finalised to te calibration of te sum of weigts of te different levels of te population and to te elimination of te bias caused by different non-response rates among strata. In our specific case, te base weigts are: v i r s r LFT n LFT i 38. Te ypotesis underlying tis metod is tat te total number of non-responses does not ave an impact for omogeneous groups of statistical units. 39. It is demonstrated tat, for omogeneous groups of responses (response omogeneity group RHG), te estimator is unbiased. ˆ n ( r), RHG i 1 Y v i y ( r) i 2 For greater detail, te reader is invited to refer to IREPA: Osservatorio economico sulle strutture produttive della pesca marittima in Italia. 2000. Ed. F. Angeli, metodological annex 8
Non-sampling errors 40. Te final step of te survey is te ceck on te elementary data to eliminate non-sampling errors. Tis as been acieved by means of computer programmes implemented to correct te erroneous values and to permit statistical data analysis. Tese programmes are mainly based on grapical analysis of elementary data. 41. Oter standard interconnected computer programmes were added to support procedures for controlling, filing, correcting and elaborating data. Tese are able to facilitate te process of assessment, transmission and diffusion of statistical information. 3. RICA SOFTWARE 42. Specific software as been developed to conduct te survey. 43. Te software is divided into te following sections: Data collectors software to be filled and transmitted. Software for data processing: queries on specific groups of elementary data. Software for te production of te final tables: cecking and correction of te elementary data, application of te expansion factors. 44. Software is developed on Delpi language. Data bank is structured on Interbase module. 45. Specific statistical software (Windows Statsoft) is used to treat and to analyse data for scientific purposes. 9