The state of agricultural sectormodelling and
possibilities of further development basing on similarity

BUNKÓCZI, LÁSZLÓ, Pitlik, László

Institute of Methodology, St.StephensUniversity, H-2103 Gödöllő, Hungary

Keywords: sector modelling, simulation, automation, quality assurance, similarity analysis

ABSTRACT

Since 1980 the regional agricultural analyses basing to statistical data assets (with doubtful authenticity at national level also) were raised to formal grade within the EU, and from these the most voluminous and complex version is the agricultural sector modelling. In the next pages a “critical” interpretation of an EU-project (CAPRI[1]: Common Agricultural Policy - Regional Impact Analysis) - closed in the beginning of 2007 - and the usage possibilities of some results from the general modelling and futurology at sector modelling will be presented.

The aim of the study is to point out, that only in that case, when the idea of consistency is well and detailed drafted can the modelling awaited to increase it’s authenticity and the efficiency of it’s construction, as without these the social benefit of the models - because of the relative high labour need and the uncheckable accuracy problems - can be questioned…

introduction

Justification of the topic choose (motivation)

The authors in the last decade took part in the largest (at national and EU level too) agricultural sector modelling projects in basing and testing (SPELGR, PIT, IDARA, CAPRI – cf.: The model-development steps themselves are connected to the Institute of Agricultural Policy at the BonnUniversity (IAP - and to their projects. The basing process can be seen as constructing the Economic Accounts of the Agriculture (EAA) and national and regional level forecasts (e.g.: expected yields) both defined as model input. In the frames of the testing process the task was to find arguments for and against (cf. with consistency criteria) validating the datasets outcome as result of the model calculations.

Other researches (connected to Prime Minister’s Office and to PhD works) running parallel with sector modelling approached and examined the anomalies of data assets management and the possibilities of the automation of modelling (OTKA F030664, T049013).

The occasion for publishing the experiences and recognitions gathered along years is the closing of the CAPRI project, as all project-closing equals with the appointment of new research directions also.

The aims of László Bunkóczi`s PhD research is to explore the anomalies of agricultural sector modelling, to draw solution suggestions, to change the present rather intuitive based forecasting practise with a consistent, plausible and checkable (either dynamic) future generating, and thus in this study his task is to evaluate the CAPRI model.

In the centre of László Pitlik`s researches there is the exploration of the possibilities of automatic knowledge acquisition basing on measured data. Testing the recognised general knowledge affects the agricultural sector modelling also, as the most complex consistency idea is drawn within the questions of the economics of the agriculture up till now. The suggested similarity based analyses gives the methodological frames to automate the present intuitive forecasting practise.

Short overview of the agricultural sector modelling (source: How the EU`s agricultural policy is made, planned University note, made by the support of DAAD:

-SIMONA: Behind the abbreviation there is the simulation and monitoring model of the former German Democratic Republic (GDR). The model was born before the joining process of the former GDR to the EU, after the assignment of the German allied agricultural government and had aims in agricultural sector modelling.

-QUISS: The model constructed in the seventies was known as a pioneer project of the German modelling, and that’s why it was better used in a groping and in an experimental way rather than a device for direct economy political consultancy. The obtained experiences were built into the latter SPEL, RAUMIS, and CAPRI applications. The aim of the model was to make quantitative analyses in regional level and by farm types and to give information about the agriculture.

-DAPS: The predecessor of the latter SPEL and CAPRI models. Sectoral like (not regional). In it’s name there are hints for dynamic analyses and for it’s forecasting system.

-RAUMIS: The regional models (RAUMIS model family) have a higher resolution than the EU and World trade (WATSIM model), and these models supply the data, for more detailed (farm group level) analyses (DIES). Demo data for RAUMIS (from 1991) can be reached under this URL:

-DIES: The DIES model basis on the data of the Farm Accountancy Data Network (basically accountancy), but functionally in it’s methodology it concerns to the EAA based solutions.

-PIT: In the frame of the PIT project, the first not formal, but SPEL methodology based database was created between 1997 and 1999 for the assignment of the EU, and on the base of it the ex-post analyses of the Hungarian agricultural development was made (Köckler, 1999). The PIT database is the starting point for the simulation module of the IDARA project.

-SPEL – Sektorales Produktions und Einkommens des Landwirtschaft (1996-1999)

-IDARA – Integrated Development of Agricultural and Rural Areas (2000-2003)

-CAPRI – Common Agricultural Policy - Regional Impact Analysis (2003-2007)

Material and methodology

The base material for the critical analysis is from the CAPRI - closing in the beginning of 2007- project’s databases, model runs and methodology. For the analysing of the possibilities of automating partially the sector models, the experiences of the case studies - from wider and wider area (more than hundred independent area) - of an own developed methodology were used. The process forming these new similarity based solutions for online service is supported by the INNOCSEKK 156/2006 project (c.f. my-x.hu).

Data content: The „material” of the analyses is the data. For the first step it’s necessary to overview the basics of data assets management from agricultural sector modelling:

What kind of data assets is handled by a national level agricultural sector-model (cf. SPEL)?

-For starting database it’s perfect, quite good as an example and makes possible the settlement in the direction of a column (sectoral) and in a row (national product production, consumption, export import).

-More dozen plants and about one dozen animal husbandry sector or main product.

-Decades long time series (averages) for the countries of the EU and from other connecting countries.

There are four kinds of model (cf. SPEL):

-Short time forecasting model / simulation: After the central hypotheses of the SFSS, the most important decisions concerning the production are already made. The production is in progress. Thus the most important exogenous variables can be defined (estimated) by experts (not on a statistical base). Only the substitutable resources and products (diverse fodders) have to be calculated by the model in an endogenous way.

-Middle long forecasting model / simulation – simple: The MFSS1 is a device for political analyses and serves as a support tool for decision preparation along dialogues between politicians and decision preparatory. MFSS1

-Middle long forecasting model / simulation „complicated”: MFSS2

What contains a regional model (CAPRI vs. SPEL)?

-COCO (Consistent and Complete) starting database for each present EU member state

-for their NUTS II regions and for more than 1.000 farm types (CAPREG), and

-for large regions of the World,

-modelling environment load for:

  • fertilizers
  • manure
  • pesticides

-Age divided groups for animals

  • Egg. bull/cow, heifer/young bull
  • Diverse feed demand, diverse manure,

-general welfare indices (GDP, GNP), and

Used data sources: EAA and FADN

The limits of the data assets:

-Using FADN data, theoretically it would be possible to run a simulation for a specified single farm – supposing, that an FADN farm is fully representative – but it’s not. Because of the earlier item had to create farm types by NUTS II regions, which behaves almost the same way, has almost the same size, with similar activities and quantity of outputs etc. In case of certain regions 10 farm types aren’t enough but at another place 3 is enough.

-After the datasets supplied by the national statistical offices it’s not possible to create a consistent area balance of a country, that could be used the clearly see the inner logic of a country’s, region’s land use. The 5-10% error only for the arable land questions what we optimise in the end. When the land as a limited resource can’t be settled as basing a model, how it can be expected, that the animal stocks and trade strategies built on the ratio of sown area from a model, to be realistic? When it’s about a relative small sector, the more dangerous is the disturbance caused by the gross area…

The inner model of CAPRI

There’s only one model/scenario in a highlighted position:

-BASELINE or standard run for 2013.

Beside it there are countless possibilities with the:

-Scenario databases and program files.

Comparing to SPEL here there’s no short-, mid- or long time model/forecast. Here is given a forecast/model run for 8 years in the future, which is made under the „ceteris paribus” principle where the environment and all the affecting factors are constant. Here come the scenario runs where it’s possible to make a model run with different settings/factors which could give answer for that question what would happen if something new measure is introduced (e.g.: customs) or the way of subsidy were changed. The result is the difference between the constant and planned agricultural policies, which in this form, in the absence of future statistical data can never been checked.

The principles of the CAPRI model:

a 3 year average – or an average year - is created from the starting database (2002-2004),

a future state is created for 8-years, which can be interpreted as the average of the target 3 year (minimising the impacts of weather),

expert opinions are used basing on future linear trends visualised to define the future yields, the inner/international supply and demand,

along creating the future state, the goal (function-?) is always to maximize the income of the given region,

among affecting factors the inner and outer supply and demand are taken into account,

for calculating prices an iterative method is used which supposes that there is/will be always equation on the market between Supply and Demand,

contains constraints, which has to be taken in a forced way into account by the model, not to reach extreme results (market balance, production, prices, production value, consumers behaviour).

After this description it can be summarized, that in case of all plants the crossing point of the supply and demand quantities - curves (?) - is made in an iterative way where the curves/functions are made by linear trends, and the goal is among all steps to maximize the region’s income taking into account the constraints.

Other modules of CAPRI:

-Complex modelling of husbandry:

  • Divided age group of animals,
  • Diverse feed demand and environment load by species, for genders and age groups,
  • Remark about the quality of data: On the base of the regional pig substance statistics the authors didn’t manage to find that kind of parameter combination to validate all the biologic and technologic expectations in all the counties and to deduce the national indexes from the indexes of counties (e.g. mass of cutting) (

-Model for environmental load:

  • Nascent ammonia, methane, carbon dioxide, N2O,
  • nascent manure,
  • used fertiliser.

-Modelling indices reflecting general developed state of the economy and society:

  • GDP,
  • GNP.

Software background:

-All the routines are executed under the GAMS program language. The extension is *.gms.

-Storing the data is under the GAMS data exchange format: *.gdx – freeware

-User interface:

  • Java based User Interface(s)
  • Results under XML base under Explorer – „old”, or in the „new” version:
  • Table
  • Figure
  • Map

reSULTS

1. The critical characterization of sector modelling:

The tasks of the initialisation and testing of sector modelling are to ask for the principles of data assets management and to ensure the interpretation of the model run results.

Partial examination of THE results`s of a capri model run concerning hungary

Examining the values for 2013, the following part of the table which is highlighted for the first sight:

Sector / 2002
(1.000 ha) / 2013
(1.000 ha) / change
(%)
Other oilseeds / 37,87 / 1,87 / -95,06%
Pulses / 27,84 / 0,12 / -99,59%
Sugar beet / 57,88 / 246,98 / 326,75%
Other industrial plants / 2,68 / 0,06 / -97,66%
All land cultivated / 6.886,71 / 6.627,8 / -3,76%

1. table

Production areas and their changes 2002-2013 CAPRI model run (baseline) results (date: 27.10.2006. work material)

After this table, assuming unchained agricultural policy, for the average year of 2013 on the base of the 2002 year, there will be 246,88 thousand hectares of sugar beet instead of 57,88 thousand hectare in Hungary beside unchained population and market tendencies. Of course starting out from the average year of 2002, it neither can be taken into account, that in 2006 the sugar factories in Kaba and in Kaposvár, before it in 2005 in Hatvan and before that in Selyp was closed, and nor that the EU Commission in the autumn of 2006 began again reviewing the necessity of the sugar beet based sugar production within the EU, instead of the higher sugar contained sugar reed, though that shall be imported. Not taking into account the topical events - as a BASELINE run shall not know them – though raises the question: if the production area is forecasted by trends, how it can be, that the value is positioned into that height which is more times higher than the chronological maximum, without specifying the demographic and market demand which could impact this huge increase.

Also conspicuous from the listed numbers the 95,06% decrease of the other oils seeds in a period, when the renewable energy resources can be identified as general strategical aims. After the information of the moment of the analyses: the government could agree with the „national” oil company (MOL), and it seems that the factory in the region of Nagykunság - delivered in 2002 - now – after 5 years - will be able to start with bio-diesel production (extruding and etherisation) what is made from rape seeds. Reflecting to this agreement this 95% decrease is unintelligible, where the latest information means only the realisation of the known strategy, and not a cardinal strategy change. There’s the question again: how this minimal value under the chronological minimum (1.870 ha) can be deducted and what kind of powers lead to this point.

Other significant industrial product’s production area - like the maize - doesn’t change in a significant measure after CAPRI BASELINE run. Starting out from the chronological data a significant change is not necessary also, and examining the bio-alcoholic project’s plans, perhaps the reduction of the area of maize won’t be not needed to reduce, but this isn’t sure for the wheat.

Not referencing for the topical-political plans which affect the agriculture’s future, it’s important to see clearly, that in case of trend based forecasts using the whole chronological time series it’s hardly likely to predict 320% growth or 95% decrease. Whether only the data of the last three years are used for an 8-year long forecast, than a drastical 20% year/year change (along 3 years) may cause that less than 100% decrease or more than 100% increase. Thus from this point of view it wouldn’t be allowable to work only with that 3 years long „time series”.

As the length of the period for forecasting was given as a subjective expert opinion, arises the question logically, how can the given data (all time series) be processed in a unified methodology, in that way that the expectable change of the production levels (as real endogenous variables) shall come into being in the form of the most consistent estimation. This will be described in a separate chapter and the methodological general potential of the similarity analyses will be presented.

The 4% reduction of all cultivated area can’t be said to be significant, but in this case we should hold in front of our eyes the principles of balanced statements and have to specify also where this out falling area appears and why do we have to count with out falling areas from an economical point of view.

2. The possibilities of utilisation of the similarity analysis for improving the authenticity and the efficiency of sector modelling:

The similarity analysis always starts out from an object-attribute matrix. Anything can be an object, or an attribute (X, Y). Along the analysis the next general questions can be answered:

-can it be shown out, and whether yes, in case of which objects there is illogical state (inconsistency) by the examined phenomena (Y)?

-for which kind of rule set (expert system) can the relation set between X and Y attributes inducted (cf. ceteris paribus, yield function)?

-how figures the value of Y (simulated, and forecasted) on the base of arbitrary X-es?

-which X-es are in tight relation with Y, and which attributes becomes noise (perturbation) on the base of the analyses?