Selection of sheep husbandry technologies under single and multiple goal constraints

Introduction

Agricultural development involves the introduction of new technology that, at least potentially, should increase production and profits. The development process itself creates new conditions which affect the choice of technology. As development usually affects a whole region with all its interactions between land, labour, capital and climate, the course of technology diffusion can be very difficult to predict. This is so when one goal, e.g. profit maximisation, is dominant. It is even more difficult when different participants in the development process have different goals that sometimes conflict with each other. A single participant can also have a number of goals that may conflict, e.g. income stability and income maximsation. A method that could indicate likely paths of development with the complementary technology selection would be useful in formulating both development and research policy. In this paper we will deal first with the single goal situation. The multiple goal approach will then be discussed.

In the single goal situation, development is seen as a series of technologies that are selected in order to maximise regional consumption over an extended period. The approach adopted includes the use of a multi-period linear-programming model that allows comparisons of many technological options and considers the economic and natural factors that determine the rate of technology adoption. The existing options open to the farmer and the resources of the region have to be defined. The abundance of specific plant and animal genetic resources is of particular importance to agricultural production, which may be limited by biological reproduction rates, especially in the early stages of introduction.

The analysis can be conducted for a number of future scenarios. The questions of interest would include the following: what mix of agricultural technologies, new and old, would tend to be selected under given regional conditions, prices and policies, over a period of time? What is the marginal benefit of a particular new technology? Which technologies have no significant value in the regional context and need not be considered in a research or development scheme? .

Definitions

The following definitions have been used in this study.

A production system is a set of components organised and operated to convert inputs into useful outputs (Salter, 1966). The components of the system are units of hardware (e.g. machinery, fertilizers, genetic stock) with appropriate software (operating instructions). These components incorporate information (know-how) that has been obtained by experience or by research and development (R and D). Technology is the science that deals with the information on which the system and its components are based. In the present paper, and elsewhere, technology is used in a narrower sense and refers to the specific information content of a given system. A 'new' technology is derived from an existing one by changing at least one unit of information. Similarly, a new system is derived from an existing one by changing at least one component. An activity is a description of a production system in terms of its input:output ratios expressed in physical or money units.

A region is defined by its borders, resources and distance from major trade centres. Theoretically, any product can be transferred through regional borders; in reality, what can be transferred is determined by political and economic considerations and sometimes by veterinary and similar restrictions. In our study, it is assumed that concentrate feed, fertilizers and mutton can be freely transferred but there is a restriction on the transfer of labour, capital, land, ewes and roughage produced locally as a byproduct of the cultivation of wheat or other crops.

Regional resources (or constraints) are the amount of available labour; cultivable land area and range are, which with their specific climate determine primary production potential: initial number of ewes of different breeds and crossbreeds; and the initial amount of physical capital.

Agricultural system inputs include hardware (physical inanimate inputs, e.g. sheds, equipment, fertilizers, concentrate feed), software (a set of operating instructions, e.g. feeding and grazing regimes), and genetic stock (animal and plant material) which is a category of hardware specific to agricultural (and other biological) production systems. Slow rates of reproduction can be an effective constraint to the adoption of any technological changes dependent on new genetic stock.

A physical, inanimate input can have a long technical life (e.g. sheds and fences) or a short technical life, i.e. it is consumed within the production year (e.g. medicines or concentrates). It can become economically obsolete before it becomes technically obsolete (i.e. physically unusable). Software does not become technically obsolete, in the sense that it does not wear out, but it can become economically obsolete. The genetic stock also does not necessarily become obsolete as it can replace itself by reproduction. Individual sheep or flocks of a given breed can, of course, become economically obsolete. The single goal of this analysis is to maximise regional consumption. Other possible goals include maximum production, improvement of regional balance of payments and maximum employment. The goals can be treated as constraints or can be used as target functions in an iterative procedure (Nijkamp and Spronk, 1978).

System management is understood as the choice and timing of alternative paths of action available for operating a system or a set of systems. It can be divided into strategic and tactical choices. Strategic management involves choice between systems: tactical management involves choice within systems. The present study simulates strategic management within a region. It does not deal with each farmer separately, but assumes that economic opportunities that are created in a developing region will be exploited by one farmer or another.

Approach

Information on activities and regional constraints can be set up as a linear-programming (LP) model in which an optimum mix of activities over time is determined (Beneke and Winterboer, 1973). The LP routine can be applied in a multi-period format or recursively (Cocks, 1965; Storck and Shurmer, 1970; Porat, 1972). The recursive approach can be used in an interactive manner that does not assume prior knowledge of prices, constraints or climatic changes. In that sense it probably simulates farmers' behaviour fairly accurately (Day, 1963; Heides, 1966). In the multi-period approach (MPLP) the target function can be optimised over the whole planning horizon. It uses average input/output values and so does not confront directly the problems of climatic uncertainty and risk that characterise most agricultural regions. It assumes prior knowledge of prices, constraints and climatic changes.

In the present study, only the multi period-approach has been developed as it is more appropriate to problems related to the time-span of technology selection and to long-term policy assessment, in which problems of year-to-year management can be neglected. MPLP can simulate many of the important features of strategic farm management in the following manner:

1. System selection is based on long-term optimisation and is calculated for the whole planning horizon.
2. The target function can be adapted to the characteristics of a specific region. It could be maximum meat production if the development goal was to increase food supplies, maximum return on investment if the study was conducted for an investment agency, or maximum consumption if regional welfare is the major goal. We chose maximum consumption as the most appropriate target function for our specific region. Other objectives, like a minimum return on investment or social security, can be defined as constraints.
3. System selection can be implemented by transfers of various sorts from one period to the next. These transfers include investment (financial capital to physical capital), young female stock to breeding ewes or to mutton, hardware from one system to another, and transfer or sale of breeding stock from one system to another.
4. Consumption and investment: The choice between consumption and investment is made endogenously (Kislev et al, 1971). A subjective discount rate (or capitalisation rate), S, provides a common denominator for consumption over different periods. An objective discount rate, R, represents the investment opportunities outside the agricultural sector. The two discount rates, as well as the opportunities for on-farm investment, govern management decisions concerned with the amount and timing of investment and consumption.
5. The economically justifiable price that can be paid for additional  resources (e.g. imported sheep or straw) at any time is given in the model by the shadow prices of the various constraints throughout the planning horizon.
6. When implementing a new technology the farm manager will try to use as much of the existing capital assets as possible until they become technically or economically obsolete. The model simulates manager behaviour by optimising the choice among the following options: use of existing assets if the new technology allows it; postponing the acquisition of assets specific to the new technology until the old assets become obsolete: and investment in new assets to exploit new technology.
7. It is possible to adjust the herd (and breed) size and composition according to the demands of the selected technologies by retaining animals for breeding or selling them. The periodic decision to cull and sell hoggets depends on the alternative uses of annual income for consumption and investment outside the agropastoral sector as well as on the revenues expected from the larger herd. Thus, the herd population dynamics are a result of both biological and economic factors.

The model can be used to analyse different development policies as well as to assess the short- and long-term impacts of a wide variety of technologies under different socio-economic scenarios. The information generated can also be used for regional planning purposes as it optimises the mix of technologies required during each period along the course of development. However, one would then have to be sure that all the technologies in the R and D pipeline will indeed be available for implementation at the appropriate time.

Case Study

General

This analysis is applied to the specific case of agropastoral development in a semi-arid Mediterranean region in Israel where the traditional activities of sheep husbandry and dryland agriculture are now undergoing extensive change. This is due not only to research and technology transfer but also, if not mainly, to the parallel development of input industries (equipment, concentrate feeds, recycled agricultural wastes, veterinary services, etc.). These inputs increase the range of optimal systems but also increase the alternative values of labour and capital. In the present study, wheat cultivation is introduced as representative of alternative agricultural activities to sheep husbandry. Activities outside agriculture can be assessed by comparing shadow prices for labour calculated by the model with salaries in industrial and commercial activities. The agropastoral systems in this region have been surveyed and studied fairly extensively (Noy-Meir, 1975: van Keulen, 1975: Noy-Meir and Seligman, 1979: Zaban, 1981: Benjamin et al, 1982). Detailed information on the regional characteristics, existing and potential technologies and the method of system definition is given by Spharim and Seligman (1983). The sheep husbandry systems are summarised in Table 1. The coefficients for the linear-programming matrix were generated by a special program (Seligman et al, 1982).

Table 1. Sheep husbandry systems used in the model.

Notes

1. Weaning weights: 20–40 kg in Awassi systems; 40 kg in extensive GMM systems. 15–30 kg in intensive GMM systems; 15–20 kg in Finn-cross systems.

2. Sale weight of lambs: 20–45 kg in Awassi systems; 45–50 kg in GMM systems and 50 kg in Finn-cross systems.

3. Pasture fertilization: in systems with less than 90% lambing rates (systems 1–5) no fertilizer is applied to pasture. In the more intensive systems, nitrogen is applied at the rate of 50 kg N/ha; P is applied to the limited extent that annual legumes are grown.

4. Systems code is a mnemonic for breed (A,I,M,F) ; grazing system (Y,D,G); weaning and lamb-rearing system (N,E,H); and effective lambing rate (50–240%).

Technology selection

From the large number of technologies available, the most appropriate must be selected and ordered over time. The multi-period linear-programming approach (MPLP) is used to select an optimum mix of technologies in order to maximise consumption over a whole planning horizon subjects to a set of constraints. The constraints are the regional resources, e.g. land, capital, labour and genetic stock. A social constraint was included by stipulating a minimum sum of money ('social security' or unemployment allowance) for each labour unit available. The set of technologies includes those that exist in the region, those that are available but not yet applied ('on the shelf') and those that are still in the R and D pipeline. A planning horizon of 15 years was adopted, as preliminary studies showed little change in the solutions between 15 and 20 years. The technical aspects of the optimisation matrix for the specific purpose of the present study are given by Spharim and Seligman (1979).

Results

One hundred and eight technologies were analysed in terms of 17 different scenarios (Table 2). These scenarios varied with regard to available labour and land, breed and prices of lamb, concentrate feed and wheat. The standard run was chosen rather arbitrarily but is meant to approximate to the current resource situation in the region.

Table 2. Description of development scenarios used in the model.

*Additional characteristics that apply to all scenarios: 15-year run: 3% capital interest rate: 4% discount rate for consumption(s): 4% population growth rate; 7 years' technical obsolescence; 1 year software elasticity, i.e. systems need not be maintained for more than 1 year.

Rejected technologies

Activities that were not selected (or selected only occasionally) are listed in Table 3. The most extensive system—Awassi sheep, year-round grazing, normal weaning, 50% net lambing rate (AYN05)—was not consistently rejected, but the other extensive systems, which were based on only slightly higher outputs and moderate increases of feed inputs per ewe, were rejected in all scenarios tested. The more productive improved-Awassi systems on the other hand, were selected frequently over their whole range of management intensities. The only system based on this breed that was rejected involved green-season grazing combined with the lower productivity level (IGN07). Green-season grazing was selected only when the breed was managed at the limit of its productive capacity (IGN10).

Table 3. Selection of the technologies used in the model under different management systems.

Key: x = screened as inefficient.

0 = consistently rejected.

* = occasionally selected.

Blank = consistently selected.

(1) = See Table (1)

The two least productive (and most extensive) systems based on the German Mutton Merino (MDN10, MDN12) were consistently rejected. The most intensive system, based on the labour-saving artificial weaner (MGH19) was also consistently rejected; only the most intensive Finn-cross systems (FGE24, FGH24) were selected.

A pattern of rejection emerges when the systems are arranged in order of a dominant aspect of production performance, which, in the present study, is ewe fecundity. Each breed has a maximum fecundity dependent on its genetic characteristics. Below this maximum, fecundity is determined by management factors. As a result, the fecundity ranges of the breeds overlap. The technologies rejected tend to be those based on the highest and lowest fecundity levels of a given breed. At the highest levels, a more fecund breed tends to be selected, and at the lowest fecundity levels of a given breed, another less fecund breed tends to be selected. These less fecund breeds are often hardier and thrive under less favourable conditions.

Selected technologies

The degree to which a system was selected over the whole 15 year test period can be expressed as the number of ewes that were managed or the number of lambs sold in that system totalled over the whole planning horizon. Another criterion is the number of times a system was selected over the period analysed. A summary of the systems selected is given in Table 4.

Two basic technologies were consistently selected in all combinations: the improved Awassi, year-round grazing, normal weaning, 90% net lambing rate (IYN09) and the German Mutton Merino (GMM, deferred grazing, normal weaning, 140% net lambing rate (MDN14). Two other technologies (MDE16, MGE19) were selected in all but one of the combinations tested and one (FGE24) in all but two; four (AYN05, IYN07, MGN17, FGH24) were selected, but not consistently and not heavily, and one system (IGN10) was occasionally selected, sometimes with many sheep involved.

Table 4. Technologies most frequently and heavily selected in the model under different management systems.

System(1)		Range		Cultivable land				Feedlot only
No.Code		Unfenced	Fenced	Unfenced		Fenced
		minimum straw (OS)	Min conc (RC)	Min. straw (US)	Min. conc. (UC)	Min. Straw (FS)	Min. conc (FC).	Min. straw (NS)	Min cons. (NC)
1	AYN05			+	*		*	Not	defined
4	IYN07	*			*			Not	defined
6	IYN09	**	**	***	*	***	***	Not	defined
7	IGN10	*		**		+		Not	defined
10	MDN14	**	**	+	*	*	***	Not	defined
11	MDE16	*	*		*	*	***	Not	defined
12	MGN17	Not	defined			+
13	MGE19	Not	defined	*	**	*	*	*
17	FGE24	Not	defined	+		*	*	**
18	FGH24	Not	defined			+	*	+

Key: Blank = consistently rejected.

+ = selected occasionally to a moderate degree.

* = selected often to a moderate degree.

** = selected often, sometimes heavily.

*** = often selected heavily.

(1) = See Table 1.

The most heavily selected group was of the intermediate breeds, the improved Awassi and the GMM, both at a production level below the breed potential. The highly productive systems based on the Finn-cross were selected less regularly and, as will be seen below, only under special conditions. The systems that were selected only occasionally range from the most extensive to the most intensive systems defined.

Dynamic aspects of technology selection

The choice of breed and production level varied over the period analysed. In the standard run (see Table 2), the Awassi was predominant initially (the present situation in the region) but its numbers declined as GMM increased in numbers, (Figure 1) . The highly productive Finn-cross was selected to the limit of its availability up to year 9, but was then phased out. It was selected during the early period as an interim solution for increasing the more productive systems as quickly as possible, but when there were sufficient number of the preferred GMM for optimal use of the resources the more productive Finn-cross lost its relative advantage of high fecundity. A different result is also possible as this pattern is dependent on the regional constraints. If the standard-run scenario changes, and the labour available increases from 400 to 1000 man-years, then the labour-intensive Finn-cross is selected to the limit of its availability throughout the test period (Figure 1). However, this is accompanied by a drastic reduction in income per man-year.

Figure 1. Predictions of breed selection over a 15-year period

Effect of price change on technology selection

An increase in the meat:grain price ratio produced a sharp increase (from 5 to 73%) in the area of cultivable land used for pasture (Figure 2) and in the herd size, all breeds increasing as a rule. The systems based on the Awassi breed that were selected are the efficient group of land-intensive technologies, and the selected systems based on the GMM breed are the efficient group of concentrate-intensive technologies. When prices of meat rise (and prices of wheat are left unchanged) it becomes more profitable to produce meat from pasture by means of Awassi-based technologies, and also from concentrates via the GMM systems. As a result, the Awassi herd increases at the expense of land under wheat, which is put down to pasture, and the GMM increases by using more concentrates. On the other hand, as the wheat price rises, the area under wheat increases and the area of cultivable land under pasture decreases, from 61% to 29%. As a consequence, the number of Awassi sheep is reduced, but the GMM and Finn-cross numbers remain more or less constant.

Figure 2. Effect of meat: grain price ratio on razing system, land use for grazing and total flock size

The increasing meat:grain price ratio has a somewhat unexpected effect on the grazing systems that are selected: year-round grazing increases, deferred and green-season-only grazing decrease, while confinement to feed lots increases only slightly, even when the ratio is 14:1 (Figure 2). Under these circumstances it may seen surprising that range use decreases. These trends in pasture use are also related to the relative profitability of wheat in the region. When the area under wheat is reduced by higher meat prices, cultivable land that can produce good pasture becomes available for sheep husbandry, thus providing an apparently more profitable alternative to raising sheep on either concentrate feed or on range. This is an interesting result for those who deal with future trends in sheep husbandry, as even under scenarios with a meat:grain price ratio of 14:1, most sheep are raised on improved grassland rather than under feedlot confinement. However, extensive rangelands tend to be neglected as meat production from improved pasture becomes more profitable per labour unit. The shift from open to fenced range and pasture occurs for apparently the same reason.

With regard to the specific technologies selected, the general trend was to select more productive systems (higher lambing rates) as the price ratio increased, but there are many exceptions: GMM at 140% lambing rate, grazing unfenced range for 9–10 months of the year, was heavily selected when the meat:grain price ratio was 6:1, while improved-Awassi grazing year-round on sown pasture with a lambing rate of only 90% was heavily selected when the price ratio was 14:1. However, this is very much dependent on the initial conditions, which stipulated a herd containing 97.8% Awassi in the first years of the analysis. In the long run, the more extensive systems, with their associated breed (Awassi), tend to be phased out.

Multiple-goal analysis

The present model is now being developed into a multiple-goal model in the manner presented by Nijkamp and Spronk (1978) . Legume pastures have been added as an additional option. Preliminary results are available but a detailed study is in preparation. A different approach to multiple-goal analysis of a non-dynamic situation has also been developed (Spharim and Seligman, 1984). This is based on an input/output model of agropastoral systems and the definition of a number of relevant regional goals. The model does not attempt to find a single optimum solution. Instead, it defines a feasible set of technologies that meet minimum threshold standards for each defined goal (Figure 3). The choice of technology within the feasible set is then dependent on other factors that vary from farmer to farmer and depend also on the relative strength of interest groups that influence the development situation. Some farmers may even practice technologies outside the 'feasible' set because of resources or goals that are not covered by the model. Consequently the model solution can be used mainly to focus on the main issues between technology and development policy.

Figure 3. The feasible technology space, S, bounded by threshold isolines of the multiple-goal indices

Acknowledgements

This study is part of a Dutch-Israeli research project, Actual and Potential Production of Semi-arid Grasslands, Phase II, supported by DGIS (The Netherlands), the Israeli Ministry of Agriculture and the Hebrew University of Jerusalem, Israel. The contribution made by colleagues affiliated with the project is gratefully acknowledged. Our special thanks are extended to Professor I. Noy-Meir, Department of Botany, The Hebrew University of Jerusalem,. who participated actively in the development of the study; to Mrs Rivka Spharim, who contributed in all phases of the economic analysis; to Mr N. Hen Porath for his expert programming; to Mr Roger Benjamin, ARO, Israel and to Mr Eugene Ungar, Department of Botany, The Hebrew University, Jerusalem, who reduced the mass of raw computer output to manageable proportions.

References

Beneke R R and Winterboer R. 1973. Linear programming applications to agriculture. Iowa State University Press, Ames, Iowa, USA.

Benjamin R W, Eyal E, Noy-Meir I and Seligman N G. 1982. Intensive ley-farming systems in the Migda Experiment Farm, northern Negev, Israel. Internal report, ARO, Bet Dagan, Israel. (Hebrew, mimeo.)

Cocks K D. 1965. Capital accumulation and Hicksian models. Farm Econ. 10:458–465.

Day R H. 1963. Recursive programming and production response. North Holland Publishers, Amsterdam.

Heides T. 1966. A recursive programming model of farm growth in northern Germany. J. Farm Econ. 48:668–684:

van Keulen H. 1975. Simulation of water use and herbage growth in arid regions. PUDOC, Wageningen, The Netherlands. 176 pp.

Kislev Y, Plasner Y and Prahia A. 1971. Multi-period linear programming with consumption functions. Center for Agricultural Research, Revohot, Israel. 38 pp. (Hebrew)

Nijkamp P and Spronk J. 1978. Interactive multiple goal programming. Reports Nos. 7803/A and 7812/F. Center for Research in Business Economics, Erasmus University, Rotterdam, The Netherlands.

Noy-Meir I. 1975.Primary and secondary production in sedentary and nomadic grazing systems in the semi-arid region:Analysis and modelling. Research report to the Ford Foundation, Grant 7/E–3, Dept of. Botany, The Hebrew University of Jerusalem, Israel.

Noy- Meir I and Seligman N G, 1979. Management of semi-arid ecosystems in Israel. In: B Walker, ed. Management of semi-arid ecosystems. Elsevier Scientific Publishers, Amsterdam, The Netherlands. pp. 113–160.

Porat Y. 1972.Perspective planning of a kibbutz farming operation with a - Multiple period LP model, Agricultural Settlement Planning Authority, Ministry of Agriculture, Tel Aviv, Israel. 61 pp. (Hebrew)

Salter W E G. 1966. Productivity and technical change. Cambridge University Press, London, UK. 220 pp.

Seligman N G, Eyal E, Benjamin R, Noy-Meir I and Spharim I. 1982. A pasture system generator. Internal report. ARO, Division of Range and Forage Crops, Bet Dagan, Israel (Mimeo).

Spharim I and Seligman N G. 1979. Long-range agricultural planning of agro-pastoral systems in a semi-arid region. Paper presented at the Conference an Operation Research in Agriculture and Water Resources (ORAGWA), Jerusalem (Mimeo).

Spharim I and Seligman N G. 1983. Identification and selection of technology for a specific agricultural region. Agric. Systems 10:99–125.

Spharim I and Seligman N G. 1984. A graphical model for relating multiple socio-economic goals to research and development objectives in agriculture. Res. Policy (in press).

Storck H and Shurmer E. 1970. Experiences with multi-periodic investment models for planning the growth of a vegetable holding. Acta Hort. 40:497–515.

Zaban H. 1981. A study to determine the optimal rainfed land-use systems in a Semi-arid region of Israel. Ph.D. thesis, Univ. of Reading, UK.

Discussion

Question – Your conclusion is that fecundity is the main characteristic determining selection or rejection of certain breeds at a certain moment. Since the prolificacy figures of the various breeds overlap, there must be other technical coefficients in the input/output relations that determine which breed is actually selected. What are those other technical coefficients?

Answer – The other technical coefficients varying among breeds are the veterinary costs and the labour requirements.

Question – An important conclusion derived from the model is that restrictions on land allocation should be lifted to improve the prospects for the development of agro-pastoral systems. Did I understand that correctly?

Answer – Yes, in some scenarios the model allocated substantial parts of cultivable land to pasture.

Question – How much does that change if you use an unsubsidised wheat price?

Answer – In that scenario the wheat acreage is reduced to almost zero.

Question – Can you please clarify what the objective function is, i.e. to maximise consumption? You are using the word consumption because you are discounting future consumption, but in fact you mean the money value of output minus operating costs.

Answer – The objective function, as defined in the paper, is maximising the amount of cash available for consumption after deductions for operating costs, investments and savings. I mentioned that this model has another version that is not covered here, in which various goals are defined. These goals are defined in accordance with the aims of various interest groups, such as regional industry, regional consumers, regional environmentalists, the settlement agency, the national government or international money lending agencies.

Question – In the agropastoral systems there is an opportunity to select between sheep breeds. Do you have a similar choice for wheat production, allowing application of different technologies? For instance, instead of ploughing and discing to use only discing or to combine the disc and the drill. or to grow wheat with or without fertilizer.

Answer – In the present model most attention was paid to sheep husbandry technologies, because that was the difficult test. The arable systems were treated rather superficially. Arable technologies in the model were continuous wheat and wheat on fallow. If the input-output relations of other technologies can be defined, however, they can be easily accommodated in the model.