A MODEL OF THE EGYPTIAN LIVESTOCK SECTOR BY William Talbot Harbaugh A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Applied Economics MONTANA STATE UNIVEH.Sl'fY Bo2:eman, Hontana April 1986 ( ii APPROVAL of a thesis submitted by William Talbot Harbaugh This thesis has been read by each member of the ·thesis committee and has been found ·to be satisfactory regarding content, English usage, format, ci ta·tions, bibliographic style, and consistency, and is ready for submission to the College of Graduate Studies. ----*~£- Date ~e~/-~~---------Graduate Commi tt.ee \.ina~rmat£1.", Approved for the Major Department. . - "za..L_z__?_t~~:__________ _ Headj Major Department:. Approved for the College of Graduate S·t.udies Date Graduate Dean I iii STATEMEN'l' OF PERMISSION TO USE In presenting this thesis in partial fulfillment of t~he requirements for a master's degree at t1ontana Sta·t.e University, I agree that the Library shall make i t available to borrowers under rules of the Library. Brief quo·t.ations from this thesis are allowable wit.hout. special permission, provided that accura·t.e acknowledgement of the source is made. Permission for extensive quota·t.ion from or reproduc·t.ion of this thesis. may be gran·t.ed by my major professor, or in ' ; ~! his absence, by the Director of Libra·ries, when, in ·the opinion of either, the proposed use of the material is for scl1olarly purposes. Any copying or use of the material in this thesis for financial gain shall not be allowed without, my written permission. Signature . frv I fr ~£ 7h~------­ Date _ _w_v~· iv ABSTRACT The Egyptian livestock sector is currently undergoing rapid changes in government policies and in demand for meat.. In order to exam~ne the implica·tions of these changes for meat supplies and prices, and.to evaluate alternative policies for the future, a quadratic prograwning model of supply and demand for the sector is developed. This model is used to examine,policies, concentrating on meat and feed t imports and feed ration distribution. I conclude. that. only large increases in imports can prevent decreases in per capita meat consumption. v TABLE OF CONTI£NTS Page 1. INTRODUCTION . . . . . . . . . . . . . . ·. . . . . . . . . . . . . . . . . . . . . . . . . 1 2. LITERATURE REVIEW.................................. 4 Theoretical Literature............................. Egyptian Literature . . . . . . . . . . . . . . . . . . ·. . . . . . . . . . . . . . 4 11 3. DESCRIPTION OF THE MODEL ............ ·. . . . . . . . . . . . . . . 13 The Matrix of Technical Coe:fficients . . . . . . . . . . . . . . . Crop Production Activities . . . . . . . . . . . . . . . . . . . . . . . Livestock Activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Milk Activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .·... Feedlot Activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Government Feed Activities ......... : . . . . . . . . . . . . . Feed and Nutrient Balance Equa·tions .. :. . . . . . . . . . . Labor Balances. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Milk Balances .............. ·.. . . . . • . . . . . . . . . . . . . . . 13 16 18 22 23 23 25 25 26 Demand . . . . . . . . . . . . . . . . . . . . . . . . . . . . ~ . . . . . . . . . . . . . . . . 27 4. RESULTS AND CONCLUSION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Sensitivity Analysis...... . . . . . . . . . . . . . . . . . . . . . . . . . Interpretation of Base Sensitivity Runs .......... 34 37 Scenarios.......................................... 1977 Runs........................................ 1990 Runs......................................... 2000 Runs ....... ·. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 39 45 50 Conclusion.......... . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . 50 5. REFERENCES CITED. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 APPENDICES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 vi TABLE OF CONTENTS (cont'd) Page APPENDIX A _, Crop and Nu-trient Coefficien·ts. . . . . . . Objective Fu~ction 57 Coefficients . . . . . . . . . . . . . . . . . . . 57 Labor Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ._. . 58 Land Availability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 Crop and Feed Produc·tion Coefficients. . . . . . . . . . . . . 61 Selling Prices. . . . . . . . . . . . . . • . . . . . . . . . . . . . . . . . . . . . 61 Nutrients From Feeds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 APPENDIX B- Livestock Coefficients . . . . . . . . . . . . . . . 63 Requirements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 Herd Parameters. . Literature: . . . . Herd Structure. Death Losses. . . Growth Rat:.es. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 . 65 . 66 . 67 . . 69 Milk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .·. . . . . . . . . . Milk Production Coefficien·ts. . . . . . . . . . . . . . . . . . . . Milk Prices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ghee Production Coefficients . . . . . . . . . . . . . . . . . . . . Milk Labor Requirements . . . . . . . . . . . . . . . . . . . . . . . . . 69 69 70 71 72 APPENDIX C Labor Coeff icient,s. . . . . . . . . . . . . . . . . . . 73 Quantities of Labor on the Farm . . . . . . . . . . . . . . . . . . . _73 Wages for Hired Labor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 Transfer Coefficients for Labor . . . . . . . . . . . . . . . . . . . 74 ~-·........... 76 APPENDIX D - Computer Program . . . . . . . . vii LIST OF TABLES Page 1. Results of Sensitivity Test . . . . . . . . . . . . . . . . . . . . . . . . . 37 2. Parameters for 1977 Runs.. . . . . . . . . . . . . . . . . . . . . . . . . . . 40 3. Results of 1977 Runs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 4. Parameters for 1990 Runs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 5. Results of 1990 Runs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ., 4 7 6. Monthly Crop Labor 7. Cropping Areas. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 8. Crop and Feed Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . 61 9. Selling Prices for Crops . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 Requirement~s. ' . . . . . . . . . . . . . . . . . . . . 59 10. Feed Nutrients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 11. Milk Coefficients.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 12. Dairy Product Prices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 13. Male Labor Costs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 4 1 INTRODUCTION The Egyptian livestock sector seems set to undergo a , I ~ period of rapid change, due to both exogenous even·ts a:nd alterations in ~olicies. The purpose of .this thesis is to develop a method for modeling the effects of some of the more quantifiable and s_ignificant of ·these changes. I am concerned with two broad categories of events, divided by whether their primary impact is on supply or demand. The Egyptian governmen·t is heavily :.involved in the agricultural sector, and most of my concern in the category of supply involves the impact of H~s policies. In demand, the most important changes are exogenous to the sector, occuring as rising population and incomes increase the demand for meat products. In order to examine bo·th categories of changes in context, a model of both supply and demand had to be developed, where equilibrium is de-termined by simultaneous solutions of the supply and demand equa·tions of the main products of the sector. The demand side of the model consists of an estimate of the demand curves for the main red meat p:t;:oduc·ts (and for imported frozen·meat}. Prices and incomes are the variables in these estimates. Because these are aggregate demand ( 2 estimates, population as well as income act as shif-ters. The supply equations come from a linear programming model of livestock produc·tion. Inpu·ts include feeds,labor, and breeding animals, and outputs are the various types of meat, milk, and milk products. In a purely competitive profit maximizing world accurate specification of the technical coefficients of production and costs and prices would ensure th~t accurate marginal cbst curves could be derived from the linear programming model. Because such a world does not apply, and because accurate specification of the technical coefficients is unlikely, many restrictions and modifications were imposed on the model in order to force it to behave in a way more akin ·t.o reality. On the demand side, I in·tend to analyze the impac·t of shifts in demand caused by changing population and incomes. Because the price increases caused bJr the. shifts in d~mcind will fall most heavily on the poor, ·t.he analysis will attempt to include the effects on the poor. The rationing of frozen meat imports is also a matter of concern primarily to the poor, and it will be considered as a means of increasing their consumption levels. On the supply side of the livestock sector, the questions of interest deal mainly with gov.srnment.al policy . . The government determines the pricing, rat.ioning and composition of the "unified feed". I Hill examine the '\ 3 effects of a more rational method for formulating ·this feed. As currently distributed, the feed acts as an incentiv~ for fattening animals, because feedlots can buy the feed at low rates and then sell it at. far higher ones on _the gray market. I will look at wha·t. might. happen if it l-Jere made available to all producers, ei·t.her at. ·the subsidized price or to the highest bidders. The government also affects livestock production through it.s widespread involvement in the determina·t.ion of cropping pat·t.erns. This involvemen·t. consists of acreage requirements, :r:o·tation requiremen·t.s, and distortionary pricing. Recently there has been interest in moving prices closer to their market levels and eliminating the acreage requiremen-ts, in order ·to reduce the implicit taxation of agriculture and also to reduce the delitereous effects of market· distortions. The main concern of ·those working towards this goal has been the effects on crops, I will look at wha·t effect these changes could have on livestock. 4 LITERATURE REVIE~'l THEORETICAL LITERATURE The theory for the model used in this thesis comes from two historical sources. The first. of these can be seen as having its origins in the "Tableau Economique" of Quesnay, and consists of attempts to model ·the production side of an economy by a series of coeffici~nts representing inputs to and outputs from various sectors. This idea eventually evolved into Walras' attempts ·to mathematically describe general equilbria, and then Leontief' s system of inpu·tou·tput coefficients. Activity analysis in linear programming is a relat.ed idea, insof~r as it also relies on fixed coefficient constant returns to scale assumptions about production functions. The historical path of the second part of the theory is more modern, origina·ting with a 1951 paper by Stephe_n Enke. This stream of the li·terature is more ' concerned with computational aspects. Enke's paper "Equilibrium Among Spatially Separate Harkets: Solu·tion by 5 Electric Analogue", Enke(1951), was the starting point for the development of programming models ~Jhich allowed the simulation of market equilibria for economies which faced downward sloping demand curves. To ease the exposition I will begin with Enke. His paper was concerned with simulating the solution to a spa·t.ial market problem where a number of local markets, each initially in equilibriu1.11 ~dthout trade, were allowed to trade. What prices and quan·t.i ties trould prevail once trade was allowed, and what would the flow of goods be, subject to . . transportation costs? Enke envisioned a solution to this problem yia an analog computer, using ba·tteries and resistors supplying vol·tages and currents (corresponding t.o prices and quanti ties before ·trade) t,o points on a circui·t. board. Assuming that the excess supply functions were linear, Enke could use the linear function relating the amperage ( quan·t.i ty flows) to t;.he vol·t.age drop between two points (price difference) to simulate the change in the quantity of trade that occurs as prices change. Wires would connect points on the circuit board to allow the flow of current, while batteries with their :polarities reversed to the existing voltage difference would siand in for transporta·t.ion costs, assuming ·these were ..a linear function of quantities. With all the wires, batteries and resistors connected the computer would almost instantly reach an 6 equilibrium that minimized total power loss, and then the equilibrium prices and quan·tities could be read off it using volt and amp meters. (This suggestion should be taken as a serious one, ra·ther than merely as an in·triguing method of visualizing the problem. Such analog computers were already in use by engineers for the solution df similar hydraulic problems, among others.) Samuelson, in his paper "Spatial Price Equilibrium and Linear Programming", Samuelson(1952), saw that Enke's objective function, the minimization of power loss, could be converted into a more traditional economic one, the maximization of the area between supply and demand over all the markets. He chose to call this maximand "Ne·t Social Payoff" (NSP) in order to avoid the complications involved with the welfare implications that. t.his area usually has for economists. To Samuelson (in ·this case at least) the maximiza·t.ion of this area had the purely positive implication that it corresponds to the competitive equilibrium. The linear programming problem mentioned in ·the title of the paper refered to minimizing transportation costs subject to certain requirements on the quan·ti ties of goods going to various ports. This is clearly a less general problem of the type that Enke looked at, C!,Ild Samuelson's LP . approach to Enke's question involves solving repeated LP problems, varying the quantities of goods shipped, then ( 7 minimizing the cost of those shipmen·ts, until a pattern of shipments that maximizes NSP is found. It was not un·til later that LP was seen as a way of specifying the supply functions in these equilibrium models. ... Takayama and .Judge were among t.he first t.o consider this, in "Spatial Equilibrium and Quadratic Programming", Takayama and Judge( 1964), a paper that. was primarily 'Concerned with proposing a nonite.rative procedure for solving Enke' s problem, using the me·thods of linear programming. Their approach was to use the Kuhn-Tucker conditions of non-linear programming t.o construct; a linear model equivalent to the nonlinear one. This works, in the case of linear excess supply. func·tions, because ·the KuhnTucker requirements involve the first derivatives of the objective function, which are linear because t.he "Ne·t Social Payoff" is quadratic. Takayama and Judge also pointed out that Samuelson's technique in general, i.e. the use of NSP as an objective function, is no·t limi·ted to spatial or single product markets. They also suggested that the linear programming model. need no·t explici·tly cont.ain the supply functions, they could be implicitly included by specifying production activities, resource constraints, and input prices, as with any LP problem. Duloy and Norton made several changes in this technique. Their paper, "Prices and Incomes in Linear Progranuning ( Models", Duloy and Norton(1975), was primarily concerned with development of large scale equilibrium planning models. The basis for their technique was derived from Hartin(l975), and consisted of ,segmenting the demand function and calculating the area under demand corresponding to each quantity. Selling activities were then added to the model, with objective function values correspondin~ to the area under demand at each quanti-ty. Decreasing the size of t-he segments allowed for an arbitrarily close approximation to the shape of the demand curve, and of the area under demand at any given quantity. (An advantage of this.technique over Quadratic Programming is that demand functions need not be linear.) Duloy and Norton's main improvement was a crude technique for handling interdependence in demand. The original scheme was appropriate for more than one good only if their demands could be considered as independent. The most obvious way of dealing wit.h in·terdependent demand would be by specifing a large number of poin·ts on the demand. surface, and adding a selling activity with appropriate quanti ties of each good in the quantity balance rows and ·the sum of the areas under demand associcrted with those quantities in the objective function. For more than a few goods the number of activities needed to g~t a reasonable . approximation of demand would be enormous, and so Duloy and Norton rejected this approach. Instead they divided final 9 ou·t.put in·to groups of goods, allo\"ring zero substi·t.u·tion between groups, and substi tu·tion among the goods in a group according to the price ratios between ·the goods. The older stream of li·tera·t:.ure is concerned less with computational considerations t.han wi.t.h the question of the existence of equilibria and methods for describing the I structures of an economy, par·ticular ly the production side. Most of my discussion of this school comes from Dorfman, Samuelson, and Solow(1958). Quasnay,s Tableau was developed by Walras and Cassel into a series of four sets of equations describing the economy. Essentially this system consis·t.ed of ·. a set of functions equating ·the demand for resources in production to their supply, a set of output demand functions expressing demand as a function of prices and income, (calculated as the sum of ·the values of the resources), a set expressing the supply of resources ·to prices of output. goods and the value of the resources, and finally a set mandating the condition for competitive equilibrium, that price equal the ·cost of produc·tion. Given a· few not unreasonable assumptions, it is possible to show that is a mathematical solution to ·this set of equations, theJ.~e i.e. that this mathematical model of the economy has a determinate solution as (presumably!) the .. economy that is being modeled does. The significance of this re~ult for my purposes is that, with some not too serious adjustments, the 10 above kind o£ equilibrium modal can be related to a linear programming one, and so it can be shown that a solution exists to a programming model Hhich is also an equilibrium solution to the mathema·tical description of the economy. At this point the two s·treams of the literature can be united. The Enke-Samuelson maximand can be combined with ·the LP model to drive it towards an equilibrium solution. The result is a computational me·thod for solving the mathematical simulation of an economy or economic sec·tor. In this thesis I use a nonlinear programming approach. Although my demand functions do happen to be linear, which would allow use. of traditional Quadra·tic Programming approaches, the nonlinear programming computer program ·that I use (MINOS) is more direct. The use of nonlinear programming has de£ init.e advan·tagas. 'l'he solution to the programming problem gives useful results beyond t.he assurance that equilibrium has been at.tained. The values o£ additional resources and the costs or losses that would occur given increases or decreases in production of a .ood can be easily found from the output. of the coinpu·ter algorithm. 11 EGYPTIAN LITERATURE Various attempts have been made to describe and model the Egyptian livestock sector, although none have been discovered that use Quadratic PrC!gramming. In the latest of a series of papers Soliman and Shapouri(1985) report ebonometric estimates of demand for various types of meat. While I was unable to reconstruct their results from the data they report, .I was able toreestimate demand. functions from it t,hat served as the basis for the demand side of my model. Preston, McConnen, and Ha:rnes ( 1984) constructed a simple Linear Programming model that; served as ·the conceptual basis for the one used in this Thesis. Their model was primarily concerned with feeding a'ctivities, but. included alterna·tive ~ sources of meat, such as imported feeders and frozen mea·t. In addition their paper cons:i.dered t.he effect of changes affecting meat supply on per capita consumption, although not endogenously. In addition to these papers some reports on broader studies exist. The most relevan·t of these was found to be Winrock(1980) "Potential for On-Farm Feed Production and Utilization by the Egyp·t,ian Small li'arm Sector". This served primarily as a source of data. 12 Several papers from the Agricultural Development Systems (ADS) Egypt Project also deal ni·th the livestock sector. Again, these served primarily as data sources and are cited in the appendices. The papers by Fitch and Soliman also provided useful insight into nonquant.if iable aspects of the situation regarding meat and livestock, particularly in regards to the attitudes of government. policy makers. Cuddihiy's (1980) World Bank paper also was useful in this regard. Several sources were useful for general background information. The most interesting of these was Richard's ( 1982) book on the history of Egyptian agricul·t.ure. Ikram' s Egypt: Economic Development in a Period of Transition(1980) provided information on cropping pat.-terns and was the best. source for the informa·tion needed to make es·tima·tes about future trends in population and income. 13 . DESCRIPTION Oli' THE 1:10DEL The model used in this thesis is a Quadratic Programming model. The basic form is as follows. Max Z ,~ I = ax - (1/2 * xBx) + dy - cz, s.t. Aw <= b Where a is a vector· of demand func·tion in·t:.ercepts, x is a vector of quantities for goods with downward sloping demands, d is a vecto~ of prices for the above goods y is a .vector of quantities of goods with perfectly elastic demands, B is a matrix of slope coefficients for inverse demand functions, c is a vector of the costs of purchased inputs, z is a vector of quantities purchased A is a matrix of coefficients describing the technical requirements of production, w is a vector of quantities used in production, and b is a vector of resource constrain·ts. The purpose of this chapter is to describe these vectors and matrices and how they were derived. The Matrix of Technical Coefficients This matrix is intended t.o ac·t as a reasonable description of the production functions of the livestock sector of the economy. It also describes the crop produc·tion sector, in very abbreviated form. In ·this section I will first describe the basic ou·tlines of the matrix, then 14 describe what the coefficients represent and how ·they were obtained. Actual calculations and sources for data can be found in the appendices. The matrix consists of activities. to produce crops, to maintain livestock herds, to produce veal, fed meat and dairy products from these herds, import. activities for maize, frozen meat and feeder animals, and an activity for the mixing of the government. produced unified feed. Act-ivities to buy inputs used in produc·tion include hiring activities for labor and for supplying the nonlabor requirements for crop produc·tion. In addition there are activities for selling the various outputs. In the nonlinear part of the demand function ·t.he selling activities are for the three types of meat, cull, fed, and veal. Sales o£ milk, ghee and manure are activities with linear objective function coefficients, because no data were available to estimate demand functions. Sales of non-feed products from the crop production part of the model are also linear in the objective function, because of the .above reason as well as the fact that prices for crops are generally set by the government. The balance equations in the matrix include, beside those for the selling acti vi·ties, labor, n,utriexrt requirement.s for li_vestock, land constraints, milk balances, and balances for the transfer of feeds from crop production 15 to the livestock activities. In addition there are a large number of balance equations for what might be called intermediate ·goods. These ensure that ·the herd is large enough to provide the number of animals that are fed, that replacements are .raised . for. culled animals 1 that. raw milk makes its way from cows to ghee production and so on. The matrix is a static one. While constraints are included to ensure that. the growth of the livestock herds follows past trends, and while these constraints can easily be modified to allow any arbi·trary fu·t.ure growth rates, there is no provision for alloning ·the model to determine optimal rates of change. ( ·' The livestock production sector, including ·the production of brops, is divided irrto two typical farm sizes. One, (the first 'in the organiza·tion of ·the matrix) represents a typical small farm, of 1.2 fedqans. ( a feddan is approximately an acre) The second represents larger farms (i.e. greater than 3 feddans) whose average size is 5:8 feddans. Since the organization of t.hat part of ·the matrix devoted to each size farm is iden·tical, I wili concentra-te my explanation on the small farm. '!'he o·t.her sections of the matrix, dealing with feedlots and miscellaneous government activities, will be explained last. 16 Crop Production Activities These activities are included because positive supply functions for feed inputs cannot be es·timated. Since it seemed certain that shifts in the demand for meat and in the prices recieved for crops would affect the supply of ~e~ds, it seemed best to include some means, alb.ei t a crude one, for modeling this response. Ac·ti vit,:les are included for ·the production of those crops that take np ·the major part of t.he land base and which. provide most of tlie domestically supplied feeds. These crops are cotton, :rice, maize, and sorghum in the summer, berseem (Egyp·tian clover) and wheat in the winter. Berseem is fur·ther cl:t-·;.rided in·to four types, each allowing a different number of cuts to be taken, and one supplying seed for the following year. There are land constraints for the overall amount of land available in each season, and constraints that initially hold the land devoted to each crop at the reported level, and for later runs preven·t i t from moving more than 20% from that level. This is an a·ttempt to compensa·te for the ra·ther skimpy specification of the factors that influenbe what crops are planted. Each crop production activity has a negative objective function coefficient, which is an estimate of the costs of production excluding the land and labor costs. Each activity 1'7 includes estimates of the labor requirements for this crop, by month. The crop production activi·t;.ies then supply crops and feeds to the respective balance eqtia·tions. C:r:·ops can then be sold, and feeds transfered to livestock production. Cotton will be an example of hoH ·this works. Each feddan of cotton production has direct costs of 93.2 Egyptian pounds (L.E.), entered as a negative in the objective function. In addition that feddan takes one feddan out of the small farm summer land balance, and requires labor of various amounts from 8 of the 12 monthly .labor balances. (The treatment of labor is described in. greater detail below.) .p19 metric tons of seed co·t;.tori then are entered into the cotton balance, and .319 tons of cotton stems into their balance. The seed cotton is bought by the government, through a separate activity, and the co·tton seed cake .which remains from the cotton processing ~nter~ another balance, to be fed directly or later mixed into unified feed. Since cotton is harvested in the fourth quarter of the year the stems can be fed to lives·tock in ·that. quarter. (In reality cattle are not fed these stems, and the model ·does not allow them io be fed, but they are included here to allow cotton to have all possible attribut·es of the o·ther crops in the model.) Activities then exist that a.llow the small farmer to transfer his stems to a large farm, or to a fee~lot. There are no objective function values for any of these transfer I 18 activities, except a charge for transpor-tation, because ·the value of the feeds is de·termined endogenously. Livestock Activities There are separate sets of these activities £or Baladi cattle and for water buffalo, which combined with the two farm sizes implies fou~ sets in total. The basic purpose of these activities, besides the conversion of. feed in·to mea·t, is to require that the herd in the model has approximately the same numbers of the various types of animals as the actu~l Egyptian herds would, were they· to be producing similar amounts of meat and milk. The livestock activities allow for the quarters in which calves are born to vary, so as t,o make best use of the seasonality of feed availabilit,y. This was done because there is a definite pat·tern to births. The Winrock study reports they usually occur so that lactation will take place during the berseem season. Winrock also reported a clear seasonal pattern to growth, with mos·t taking place during the berseem season. A third reason for separating nutrient supply and demand by quarters is tha·t it reduces aggregation problems in the nutrient balan.ces. This will be discussed later. 19 The livestock activities are cooxdinated by a herd activity (HERD), which specifies the number of animals in the breeding herd and the coefficients which drive t.he rest of the activities .. · Each breeding female in, for example, the small farm cattle herd produces .632 calves. This should be read as the number of calves produced by one breeding female in one year. The .632 goes in the calf balance equa-tion. Death losses.and culling are such that .1842 of the breeding females must.be replaced yearly. The 2.6% historical growth rate of the cattle herd implies that an additional .0322 breeders will be needed anually. So the 1 in the HERD activity puts .2164 in the breeder replacement row. Each breeder animal requires .04 bulls, thls coefficient enters into a bull balance equation which -tlllsures tha·t ·the large farms have the bulls needed for breeding. The cull loss of .1842 times an average weight of 400 kg. and a 58% slaughter percentage implies that each breediug animal annually pu·ts 42.73 kg. in the cull meat balance. (See Appendix B for details) Labor requiremen·ts specify the amoun·ts of labor that men and women must supply each mon·th to m'ain·tain the herd animals. Additional coefficients represent the proportion of breeding animals on maintenance rations and a rough estimate of manure production. This HERD activity in and of itself requires no nutrients. Separate sets of ac'l:~i vi ties for calving, and 20 raising calves as feeders or breeder replacemen·ts demand ·t.:.he nutrients. These will now be described. Separa·t.e CAV activities are ~pacified for the calves born in each quarter. Each of these ac·t.i vi ties requires that. a 1 be subtrac~ed from the calf balance equa·t:.ion. Each demands nutrients from-the quarterly nutrient balances. These nutrient demands are not those of the cal£, which will be considered later, rathe.r they are ·those of the breeding female. Say we look at CAVl, a first quarter calving activity. This activity will take uut:.rie~ts from each quarterly balance in the first yea.r and the first quarter in ' the second year. The first ·two months of the calving quar·ter include nutrients demanded for gestation, the last month of that quarter and all of the two subsequent quarters include requirements for lactation. Each of the calving activities'puts a·calf, denoted by a positive one, into one of the quartorly cal£ balances. The calf from CAV1 c;:an go.from this balance into RFD1 or RBD1, to be raised as a breeder or a feeder. In addition it could be sold as veal, putting 30 kg.s into the veal meat balance, or lost to· dea'th. Each of these activities takes one animal out of the calf balance and (excep·ting VEAL), nutrien·ts from the nutrient balances. 'fhe raising ac·tivi·~.ies put. animals, via positive coefficients, into the replacement breeder or feeding animal balances. Sufficient replacement breeders are 21 required to satisfy the balance mentioned in the section describing the HERD acti vi·ty. RE'D animals enter balances from whence they go into the feedlot activities. The bulls needed for reproduction are not included in the HERD activity, rather BULL activi·ties, on the large farms, represent these animals. The ctssumption is that small farms contract with the larger ones for breeding services. A coefficient in the repl~cement breeder balance ensures that sufficient new animals are available to replace culled bulls and provide for grow~h needs. (For simplicity, these replacements. are identical to the r~;placemen·t breeder animals, although in reality they would have different nutrient requirements, etcetera.) These bulls have maintenance requirements met through t,he nutrient balances just as do the HERD animals, only there is no separate activity for this. Not all the .animals that are born will actually survive to be transfered from one act,ivU:;y t.o another. The Winrock study reports rather high losses, particularly among the young animals. One way of dealing ldth this fac·t would be ·to use tranfer coefficients of less than one and add the nutrients consumed by those ·that, die ·to the nutrient requirements of the living. I adopted the ..follmdng more complicated technique so tha·t it would be easier ·to change assumptions about death ra·t.es. 22 Each animal raised as a feeder or a breeder replacemerrt puts a negative fraction into a death loss balance for animals of that type, born in that p<'J.rticular quarter. This fraction represents the number of animals of that type that can be expected to die before maturity. Loss activities put positive l's into these balances, and sub·t.rac·t. nutrien·ts from the nutrient balances. These nutrient coef:ficien·t.s were calculated from the Winrock data on losses, and represe~t a 'typical' dead .animal, where ·the nut.rien·ts required in each quarter are reduced by the proportion of animals tha·t die before culling that have died by ·tha·t quarter. I.e. , if 30% of the animals that die do so before t.he 2nd quarter, the second quarter nutrient requirement \·dll be 70% of that for a live animal, and so on for subsequent, quarters. Milk Activities Milk from the herd can either be sold raw or processed into ghee. 'fhe ghee and the remaining skim milk can then be sold. These GE activities are divided by mont~s and by the source of the milk (cattle or buffalo). Each draws milk from the appropriate quarterly balance, and female labor from ·the monthly female labor balances. They supply ghee and skim milk to balances that allow these products to be sold. 23 Feedlot Activities Feedlots take animals out of the feeder balances and feed them to achieve various gain rates, demanding nutrients from a set of feedlot nutrient balru1ces. The feeding period is two quarters, gain rates range f:rorn . 8 to 1.1 kg. per day. Each feeding activity supplies meat. , quanti·ty depending upon the rate at which the animal was fed, to ·the fed and cull meat balances. The assump·tion is that the second quality meat from butchering is counted as cull in the government data. There are separa·te se·ts of feeding activites for buffalo and ca·t.tle (buffalo enter ·the feedlo·t.s at higher weights), but no dis·t.inction is made in the meat: it all enters the same fed and cull balances. Feeding an animal allows the I p~rchase of a certain amount of unified feed, this can then be fed or sold to the farms. Government Feed Activities '!'he Egyptian government is involved in the animal feed business in several ways but. most importantly in mixing and rationing a subsidized feed mixture. This mixture is composed mainly of cottonseed cake, wheat bran, and imported .... maize. The governmen·t obtains the co Vtonseed as residue from its monopsonistic purchases of cotton, the wheat bran from 24 milling domestic and imported wheat, and the maize from direct imports. (In addition to the maize used for the ration, some is imported and fed directly) These ingredients and several other minor ones are theri. mixed in government mills and distributed to feedlots and to farmers enrolled in the livestock insurance program, according to the number of animals owned. The livestock insurance program, and thus ·the feed ration, are effect.ively limited to large farmers. This "unified feed" (UFD), as it is called, is not necessarily fed by those who receive it, rather there is a widespread gray market, where farmers sell their ra·tion ·to ·the feedlo·ts at prices several times ·that, charged by· the governmen·t. The model deals with unified fe0d by including importing and mixing activities. Balances allot-r the feedlots and large farms to buy feed at the government price, the quantity allowed depending on the number of animals owned. Once the feed has been bought, it can either be fed or sold to other enterprises, through more transfer act,ivities. These transfers do not have explicit objective function values, the model determines the value of UH:?J t,ransfers endogenously .. As the model was set up for the calibration runs, the proportions of feeds in the unified feed w~re set according . to the formula used by the governrnf:)nt.. In later runs this mix was varied, in order to see if a mora optimal one could be found. 25 The imported maize tha·t is not used for the unified feed can be bought from the government and fed directly. In the calibration runs, the quan·tity imported was fixed a·t ·the reported 1977 quantity. Feed and Nutrient Balance Equations There are 9 separate set,s of th0 quarterly nu·trient balances for each size farm. (4 of these, one for each quarter, comprise a set.) The first two sets are for calving, The next five are for rais.i..u.g feeders and replacement breeders. A breeder replacement that is born into say the third quarter will have its initial nutrient; requirements in the third quarter nutrient balance of the first set. Nutrient balances for the three and a half years this animal takes to reach maturity will then progress quarter by quarter and year by year ·through the set.. Requirements for bulls and for main-t:.aining breeders t,hat are not calving share the last of ·these sets of balances. The last two sets of nutrient balances allow transfers to feedlots and the other size farm, respectively. Labor Balances There are separate labor balances for every morrth and for two types of labor, male and female/child. Each farm 26 size has a certain amount o£ labor available from family workers. Additional labor can be hired by the month. It is possible to transfer male labor ·to the female balances, and vice versa. One day of female labor transfers to th~ male balances as hal£. a unit of male labor. Tl1is reflects the fact that female wage ra·tes are half ·tha·t of male. On the other hand, male labor transfers to the :female balances a·t one to one, because men are not. assl\tned ·to be any more productive than women at the major :female activities, milking and processing ghee. Milk Balances Each calving activity puts milk into the quarterly milk balances. Because Cattle and Buffalo milk have different, fat contents they are kept separate. Butchering a calf as veal puts the amount of milk ·tha·t the calf would normally have consumed into the balances. From these balances milk can either be sold raw or be processed and ·then sold. Small farms consume a very large frac·tion of ·their milk produc·tion a-t home, but this was ignored, wi·th ·the effect o:f valuing this consumption at the market pricG. 27 Demand Demand for fed, cull, and veal meat was estimated using ordinary least squares on price, quantity, and expendi·ture data for 1965 thru 1977. The equations estimated were of the inverse form p = a + BQ + cY . where p = a vector of real prl.ces, Q = a vector of per capit,a quantities, y = real per capita priva·te expendi·ture, a = a vector of intercept terms, B = a matrix of quantity effects, c = a vector of income effects. The equations used were Pcull= -1.316 -.024 Qcull -.057 Qfed -.012 Qveal +.025 Exp ( 1. 0) Rsqd.= .95, D-W = 1.7, (2.8) (0.4) (16.6) S.E. of est. = .037, Mean of dep. var.=0.84 Pfed = -1.693 -.057 Qcull (2.5) Rsqd.= .97, D-W = 1.8, Mean of dep. var.=1.05 ~.073 Qfed -.045 Qveal +.033 Exp (2.0) ( 1. 2) S.E. of est. = .042, (18.2) 28 Pveal= -1.565 -.012 Qcull -.045 Qfed -.258 Qveal +.032 Exp ( 1. 3) (0.4). Rsqd.= .98, D-W = 2.0, ( 1. 7) (11.0) S.E. of est. = .041, Mean of dep. var.=1.14 where the absolute values of the T-ratios are in parenthesis, and the R squareds are adjusted for the number of independent variablep. Prices and incomes are deflated Nith 1977 as the base I I I year. In order to use these equations in the model a level of private expenditure had to be specified and they had t,o be converted from the per capita form in Nhich t.he Nere estimated to the aggrega·t.e form which the supply side of the model is in. This was done by specifying a popula-tion level and a level of· per capita expendi tu.r.·e. For each equa·tion ·the per capita expenditure was then mul·t.iplied by its coefficient and the result added td ·the intercept. The slope coefficients were then divided by ·the population ·to allow use of aggregate quantities. The above equations were t.he ones used in the model,· but ' ' others were estimated and rejected for various reasons. The demand for meat is obviously the result of a system of equations, where prices and quan·t;i ties are .. both determined by the interception of supply snd demand. Quantity is therefore not an exogenous variable, and is correlated with ' . . ' .... "··' - .,, ... ... :. .. ~ . : ':' •••• ~ .. J•• i t~..! • 29 the disturbance. The OLS es·timator will be biased. A favorite method for dealing with this situation is to use instrumental variables, or two stage leas·t squares. In these techniques one or several variables that are correlated wit.h the regressors but not with the disturbances are found. Assuming there are several of these, they are then used to calculate estimated values of the regressors. These estimated values, now ipdependent of the errors, are used as in OLS. This _qets rid of the bias problem, but unless' variables can be found that are rat.her highly correlated with the regressors the variance will increase substantially. In this case, it meant finding variabl~s highly correlated with quanti·ty, bu·t still independent. of the errors. These simply could not b€~ found. One difficulty was the nature of the livestock ~reduction process. Quan·ti ties produced are determined by a complica·ted mult.i year process. Suppose feed imports were used as an exogenous regressor. First of all, these imports are no·t really exogenous, thus they too are correlated with the errors. Even worse, they are just not very well correlated with current production. An increase in imports might; encourage more current feeding activity, or it might cause farmers to build up their herds, thus reducing curreqt production. Perhaps some structure of lagged effects could ~xplain quanti ties produced, but ·the data are inadequate to estima·t.e 30 it. Attempts were made with other, similar variables, bu-t. the results were all discouraging. There are good argumen·ts £or using OLS when only bad instrumental variables can be found. Kennedy(1979) remarks that Monte Carlo studies show that wi·th small samples OLS is less sensitive t:o errors in variables, misspeci£ica·tion, and multicollinearity. See also Mariano(1982). All the above conditions probably hold true, to some extent, in this case. Another approach considered was to es·tima·te ·the 'normal' demand equations, with quant.i·ty as 'l.~he dependent. variable. The estimated equations could then be inverted ·to give the form needed for the model. This was rejected, mainly because the available estimates of quantity were so much worse than those of price. (See below.) Errors in measuring the independent variables add to·the bias problem, because they are correlated with the disturbance term. Problems were caused by the da·ta. I had available annual price, quantity E3Xpendi ture da·ta and tHo price de£ lators, covering 1965 to 1980. Unfor-tunately one price series went. from 1965 to 1978, Hhile another covered 1970 to 1980. The two were not at all similar in the duplicated years so I used the first as it covered a longer period. The last year of this series, 1978, was an outlier. Alre~dy highly suspicous of the data, I dropped this year rath~r than allow it to distort the overall results. The deflators were 31 likewise very inconsistent, differing about 76% in 15 years., I tried using each and an average of the two, then settled on the one showing the least inflation, largely because using it to deflate expenditure produced a rate of real increase that seemed consistent with the literature's concensus. In addition to the private expenditure data, I had numbers on GNP. I considered using GNP as the income variable, on the argument that government provided services make up a substantial part. of income in a socialized count.ry such as Egypt. I eventually used private expenditure anyway, in the belief that these government serv:lces were not that important to the.demand for meat. The .~uantity measurements come from "Official slaughterhouse data" cited in Shapouri and Soliman(1985). These measurements certainly do not reflect total slaughter, I originally used the generally accepted figures that half the fed meat and 40% of th~ cull and veal meat is counted in the official figures. No sources I have found are willing to specula·t.e on whether or how these percentages have changed over time. I was unable to get the supply side of the model to produce "enough" fed cattle when using the figure that half of slaughter was reported, so I reestimated the equa·t.ions using . 66 as the proportion. These quantities agreed with the supply side of the model and with other estimates of the numbers of animals slaughtered. 32 Other difficultie:;; were caused by the kinds of meats in Egypt and the way they are supplied. Frozen meat is imported· I by the government, and during the period 1965-1977 wa~ rationed to households at .68 L.E. per Kg. This meat is not good quality to begin with, and often gets worse. I dealt with it by including ·the quantity imported as part of the supply of cull meat, because this seemed to be the only method for including it. in the demand functions. In the equations finally used, the res·t.riction was imposed that cross quantity effects be equal. It was imposed in this case to ensure that the matrix of cross price effects would be symmetric, and thus ·t.ha·t. the consumer's surplus line integral. would be unique, a requirement, tha·t ensures a unique solution for the maximization part of "tl1e model. The validity of the restriction that cross quantity effects be equal was tested using F ·t.ests. Veal-Cull rest., F = 3.3, with 1 and 24 degrees of freedom Cull-Fed rest.,, F = 1. 2, with 1 and 24 degree.s of freedom Veal-Fed rest., F = 5.1, with·l and 24 degrees of freedom The 5% confidence level for F is 4.3 and ~pe 1% level 7.8 . . These restrictions therefore s·tretch the truth somewhat. On the other hand they are necessary, and the hypo·thesis that they hold is not flagrantly disproved by the data. 33 In the above paragraphs I have described the reasons to be skeptical about the demand equations I have es·timated. In defense of the results, I would argue that the equations found seem very- plausible, and are xather robust for the number of observations and the quali·ty- of the da·ta. Nowhere do the parameters estimated differ excessively- from what economic theory- and previous work would lead one to expect,. 34 RESULTS AND CONCLUSION This chapter considers sensitivity analysis on some of the parameters, describes the various scenarios that the model is used to investigate, and repor·t.s on and analyzes the results. Sensitivity Analysis , The purpose of sensitivity analysis is to measure the "robustness" of the model. A list of coefficients that migh·t. be expected to affect the results would include the following. Coefficients of the demand equations Cropping costs and non-feed returns · Quantities of crops and feeds produced per feddan Nutrients available in feeds Proportion-of feed fed that is consumed by animals Nutrient requirements Growth rates and ages of parturition, weaning, culling Fertility rates, abortion and death losses of immature animals Quantities of milk produced Prices of dairy products and costs of producing them Weights at slaughter and ca.rcass yields Quantities of Unified Feed Ration provided Labor requirements for various activities, amounts of labor available on the farm, costs of hiring labor Returns to livestock.due to animal labor In short, virtually all the parameters in the model, plus those left out of it can change t.he results. However, 35 the effects of small changes in most of these coefficients can be expected to be small and predictable. E.g. if the nutrient requirements of buffalo turn out to be somewhat higher than estimated, we expect the model to over estima·te the number of buffalo and underestimat:.e the number of cattle. Ignoring the value of animal labor will skew the number of animals kept downward, while ignoring the energy requirements of that labor in the nutrient balances has the opposite effect~ The more important question is whe·ther small changes in coefficients of uncertain determination can cause substantial changes in the quanti ties of ·the acti vi·t.ies. The first area where this can occur is in the crop sector of the model. The major factors determining which crops are planted are the biological requirements of the ro·tation schemes and government regulations, and these fac·tors are comple·tely left out of the model. As it turns out, the model does not provide crop patterns at all similar to ·those that actually prevail and so explicit cons·traints were imposed. The other area where I felt the model was likely ·to be unstable concerned the number of calves produced annually by the breeding herd animals. and in the number of replacemen·t. breeders needed~ While changes in these nu!Jlbers would not affect the ou·tput of the lives·tock sec·tor greatiy, a small increase in the pumber for the buffalo herd might be 36 expect.ed to greatly increase the size of that herd, a·t the expense of the cattle herd. In a similar vein, changes in the relative values of other like coefficients might also be important., but since the resul-ts could be expected to mimic those for the above change thelr were ignored. As it turned out, the sizes of the herds also were very sensitive to shifts in the demand for meat, so I had to restrict these sizes at 'reasonable' levels anyway. One of the main factors determining the relative proportions of animals kep·t by farm slze is the availabili·ty of labor. Large farmers must hlre labor during much of the year, while small farmers have sufficien·t amoun·ts available within the family. The quan·tl ties of· labor available to eac·h size farm can be expected to significantly affec·t this distribution. Again, in ·the end I was forced to bind the herd levels, so it seemed pointless to pursue this, although I presumably could simulate the actual situation by experimenting with different levels of labor until I force the model to do what I want. I·t does hml'ever seem desirable to show exactly how far of£ the model is from ·simulating reality, and to this end ·two runs are made. In run A, the herd sizes and areas of crops planted.are fixed at the reported 1977 levels, in B crops are fixe4 while herd sizes are allowed to vary. For results see Table 1. 37 TABLE 1. Results of Sensitivity Test. Quantities in Thousands. ----------------------------------------------------------- ' Run: A B Quantity, Dual Activity or reduced cost in paren·t.hesis Small Farm Cattle Buffalo 751 (-121) 1180 (-70) Small Farm Cattle Buffalo 376 (-57) 393 (-60) 0 528 Meat: Cull Fed Veal 145 116 21 122 16 29 0 1279 In·terpretation of Base Sensitivity Runs It is clear from these runs that the· model unders·t.ates the profitability of livestock production. It seems likely that this is the result of ignoring the value of animal work, while implicitly including some of its costs. If this is the explanation, it is notable that, should machines come to replace animal labor meat product.ion would fall, under the assumptions ·of· this model. In reali·ty of course, ceasing the use of animals for work would presumably make them more efficient meat and milk producers, mitigating this effect and conceivably overriding it. The model apparen·tly also I overestimates the profi·tability of buffalo, compared to cattle, in fact the unres·trained run produces no ca·ttle. The general concensus, on the other hand, is that cattle are more profitable. I can offer no single explanation of this discrepancy. 38 Other results of the sensitivity runs are rather encouraging. Th~ model does produce animals, milk, etc. Run A does not result in outrageously negative returns to livestock production, and the profits from crops seem reasonable. The value of the unified feed ration is i extremely close to the actual price at which farmers resell it. · Scenarios I am interested in examining the costs and benefits of present policies and in comparing possible future policies. For the present polices, runs for the base year 1977 were made under different conditions, while 1990 and 2000 were picked as the years in which to look at future policies. It was necessary to restrict the number of scenarios, because each setting of one parameter could conceivably be examined in combination with each setting of each of the others, and the humber of combinations grows with at leas·t the square of the number of parameters of interest. There is no point in generating more information than can be digested so I eliminated broad categories of cross-combinations. For 1977, one run was made simulating the present policies and one under more market oriented alternative policies. For the year 1990 three main scenarios were examined. A examines the situation in the abscence of policy changes. B is based on 39 policy changes that are already under considera·tion, t-7hile C examines alternative ways of maintaining real prices at approximately the 1977 levels. C looks at two basic ways of doing this, one concentra·ting on imports of feeds, feeders and meat and the other on shifting domestic resources to livestock. For 2000, the initial runs showed ·that results did not alter the conclusions drawn from the 1990 runs. This is discussed in later •ections. 1977 Runs The Egyptian government bo·th subsidizes and protec·ts the livestock sector. The subsidies are mainly seen in low prices for the unified feed ra·tion, while the pro·tection results from .the effective governmen·t. monopoly on importing feeds, feeders, and meat and the willingness of the government to keep imports at low levels. To investigate the costs of these policies, two runs were done for 1977. A has the reported levels of subsidies and imports, and B has the unified feed subsidy removed and feasible levels of import.s and exports are allowed. Table 2 shows ·the parame·ters that differ between these runs, Table 3 shows the resul·ts. 40 TABLE 2. Parameters That Differ for the 1977 Runs. Parameter Bound Run A Low High Run B High Low ----------------------------------------------------------Crop Areas (Thou. of Feddans) (reported levels) Small Farms Summer: Cotton Rice Maize Sorghum remainder Winter: Berseem 1 Berseem 2 Berseem 3 Berseem 4 Wheat remainder (+ - 20% change) 658 531 933 227 592 660 533 935 229 594 527 425 747 182 474 791 637 1121 274 712 426 i93 92 703 679 547 428 195 94 705 701 549 342 155 74 563 544 438 512 233 112 845 816 658 642 518 911 221 577 644 520 913 223 579 514 415 730 178 462 772 623 1094 266 694 416 188 89 686 663 534 418 190 91 688 665 536 334 151 72 550 531 428 500 227 108 824 797 642 Large Farms Summer: Cotton Rice Maize Sorghum remainder Winter: Berseem 1 Berseem 2 Berseem 3 Berseem 4 Wheat remainder Exports of feeds: Cottonseed Cake ·Wheat Bran Rice Bran (Thou. Tons) 0 0 0 0 0 0 no limits no limits n<;> limits 41 TABLE 2. continued. Parameter Bound Run A Low High Run B Low High -------------------------------------------------------~--- Imports: (Thou. ·Tons) Maize Wheat Bran Frozen Meat Feeder Cattle (Thou. Head) Unified Feed: Rationed price (L.E./Ton) Quantity Ration Scheme 176 300 74 176 300 74 300 74 1000 300 74 0 0 0 20 25 740 Large farms, feedlots 0 n.a. 0 n.a. 42 TABLE 3. Results of 1977 Runs. Quantities in Thousands. Run: A. B Quantity, Dual Activity or reduced cost in parenthesis Small Farm Cattle Buffalo Small Farm Cattle Buffalo 751 (-94') 1180 (-96) 376 ( -107) 393 (-91) 751 (-98) 1180 (-81) 376 (-130) 393 (-143) Meat: ·Cull 145 (1.22)* 145 (1.28)* Fed 116 (1.54)* 73 (1.61)* Veal 21 (1.67)* 25 (1.70)* *These are marginal costs. For A, Veal production had an upper limit of 21, the reduced cost on veal was .42 Crops and Land: Small Farms: Summer: Land Cotton Rice Maize Sorghum remainder Winter: Land Berseem 1 Berseem 2 Berseem 3 Berseem 4 Wheat remainder 2946 660 533 934 227 592 (-98) (76) (72) (0) (-2) (-98) 2946 791 637 862 182 474 2367 426 195 92 705 681 547 (-85) (-23) 2367 (-59) 342,(-21) 155 (-8) (-4) (0) (15) (31) (-84) (-90) (114) (90) (0) (-4 (-91) 80 ( 0) 651 (0) 816 (78) 54 7 (-59-) 43 TABLE 3. continued. Run: B A Quantity, Dual Activity or reduced cost in parenthesis ---------------------------------------------------------Large Farms Summer: I ~ ~~ I I Land Cotton Rice Maize Sorghum remainder Winter: Land· Berseem 1 Berseem 2 Berseem 3 Berseem 4 Wheat remainder Unified Feed Q. Gray market value Exports of feeds: (-74) (30) (54) (-6) (0) (-74) 2875 772 623 752 266 462 (-69) (75) (80) ( 0) (4) (-70) 2310 416 188 89 . 686 665 536 (-17) (-11) (0) (0) (12) (74) (-17) 2310 334 151 73 576 797 597 (0) (-8) (-4) ( 0) 740 (4) 145 (O) (126) ( 0) 0 (79) 104 (Thou. Tons) Cottonseed Cake Wheat Bran Rice Bran Imports: 2875 645 520 911 222 577 0 0 0 112 0 231 (Thou. Tons) Maize Wheat Bran Frozen Meat · Feeder Cattle (Thou. Head) 176 (56) 300 (31) 74 0 (296) 1000 (88) 300 (88) 74 20 (320) "Social Welfare" 8.05 E+08 9.40 E+08 44 For 1977, I am primarily interes-ted in quan-tities and prices of meats, the governments cost of imports and subsidies, as domestic and foreign exchange, ·and t.he social welfare costs ·of the policies. The model echoes the results of previous studies. Government policies promote domestic n1eat production while -discouraging cheaper imports, and such policies have obvious welfare costs which are substan·t.iated by the model. ' ' I r ~ I I I ~ According to the base run, current Egyptian red meat consumption is about 7 kg. per person, annually. Without government intervention in the s~ctor (except for frozen meat imports), this would have decreased abou·t. a kilogram. On the other hand, the welfar~ cost of these policies, as measured by the area between the supply and demand functions, was about 3.5 L.E. per person, or about 3% of GNP. Expanding meat imports by 40, 000 ·t.ons would have allowed equivalent consump·t.ion levels at a cost of only about .6 L.E. per person. Both rups show evidence of the desireability of expanding imports of feeds and feeder animals: vlhile it sensible for the government to stop subsidizing the unified feed ration, feed imports are s·till a feasible proposition at market levels. Indeed, the model sugges.ts tha-t. some . exports of cottonseed cake and impor-t.s of maize t-rould provide a feed ration better than the presen·t one. See the dual activity values in Table 3. 45 1990 Runs Demand is recalculated on the assumptions that population has ~rown to 51 million and private expenditure equals 145 1977 L.E., per person. The population projection comes from the 1985 World Bank Report, while the per capi·ta private expenditure estimates are simply projections of the series used in estimating demand. The major differences in the four runs are summarized below. Table 4 gives the details, and 5 the results. A) Base Run for 1990 Herd coefficients unchanged from 1977 levels Herds have grown at historic rates and will continue to Cropping patterns unchanged from 1977 Imports of feeders and frozen meat unchanged from 1977 Imports of feeds (maize,wheatbran)_unch~nged from 19~7 B) No unforseen changes in policies. Unified Feed has urea added. 5% improvement in calving ra·t.es and death losses Future Herd growth lessens to half 1977 levels Unified feed at market price Imports of feeds (maize,wheat(bran) increase with population Imports of frozen meat increase with population Major policy changes. This includes the above changes plus CI) Imports of frozen meat increase with population Imports of feeders increase Imports of feeds (maize,wheat(bran)) increase by more than in B CD) Future Herd growth ceases A cessation of veal slaughter New cropping patterns 46 TABLE 4. Parameters that Differ for the 1990 Runs. ----------------------------------------------------------- Parameter Run A Low/High Bound or level Herd coefficients Herd growth Cat·tle Buffalo Crop areas Veal slaughter Feed Exports: (Thou. Tons) Cottonseed Cake Wheat Bran Rice Bran Imports: as 1977 3.0% 2.5% as 1977 Run B Low/High Run CI Low/High Lo~/High 5% imprvd. 5% imprvd. 5% imprvd. 1. 5% 1. 25% as 1977 free 0 0 0 0 0 0 1. 5% 1. 25% as 1977 0 0 0 0 0 0 176 300 74 229 400 0 229 400 967 0 0 0 0 0 0 0.0% 0.0% + or - 20% fixed at, 0 free free Run CD 0 0 0 0 0 0 0 0 0 0 1000 500 500 0 1020 300 400 967 300 400 967 0 0 (Thou. Tons) Maize 176 Wheat Bran 300 Frozen Meat 0 Feeder Cattle (Thou. Head) 0 Unified Feed: Rationed price (L.E./Ton) 25 Quantity 740 Ration Scheme as 1977 Mix as 1977 25 1000 as 1977 urea added 0 40 n.a. 1040 free market urea added n.a 2000 free market. urea added* *also, imported maize is substi tu·ted for cot·ton.seed cake 47 TABLE 5. Results of 1990 Runs. Quanti·ties in Thousands. Run: A B CI CD Quantity, Dual Activity or reduced cost in parenthesis Small Farm Cattle 1113(-61) Buffalo 1637(-66) Large Farm Cattle 562(..:.117) 552(,-121) Buffalo Meat: Cull Fed Veal 176(2.19) 34(2.82) 17(2.89) (-132) (-148) (-101) (-124) (-10) (-2) (-145) (-165) (-180) (-210) (-72) (-85) 198(2.18) .38(2.80) 9(2.92) 301(1.98) 111(2.50) 9(2.80) 198(2.18) 38(2.81) 0(2.96) These are margin~l cost~ in parenthesis. For A, Veal production had an upper limit of 21, the reduced cost on veal was .42 Unified Feed: Quantity 740(36) Gray market Value 55 0(79) 70 . 71 50 Exports of feeds: Cottonseed Cake 0 Wheat Bran 0 Rice Bran 0 112 0 0 0 231 0 0 0 0 2000(12) Imports: Maize 176(41) 1000(88) 742(0) Wheat Bran 300 (41) 300(88) 500(74) Frozen Meat 741 967 2000 Feeder Cattle (Thou. Head) 0(663) 20(320) 80(506) 1040(48) 999(35) 400(53) 967 0(10815) "Social Welfare", or objective function value.· 8.33+E08 9.40+E08 13.48+E08 9.67+E08 48 The results are not uriexpected. Increasing income and population will cause demand shifts that are far beyond the capacity of the livestock sector to accomodate without large price increases. If present polices continue, real meat. prices are predicted to nearly double. These increases will make imports of meat, feeds, and feeders ever more attractive, and the welfare costs of limiting these imports ever greater. Even drastic increases in frozen meat imports will not suffice to keep prices stable, as seen in run CI. Continued herd growth will consume most of the available feeds, and fed meat production will decline without substantial increases in imported feeds. Without large increases in imports, meat price icreases make domestic livestock production more attrac·tive, but this will not translate into supply increases without imported feeds. For 1990, my estimates of per capita social welfare range from 16 to 26 1977 L.E., depending on the scenario. Meat consumption ranges from 4.5 to 8.3 kilograms, while average prices go from 2. 1 to 2. 3 L. E. . Of the ·t.wo most relevant runs for policy evaluations, CI and CD, only CI, with its large increases in imports. results in an increase in meat consumption on a per capita basis. The o·ther runs result in decliries on the order of 30%. . One policy goal for the livestock sector is to increase meat consumption by the poor. I tried to present an analysis 49 of. how consumption would be affected under the various proposals, using the lowest 40% of the popula·t.ion, by expenditure, as my definition of poor. This was chosen as a definition because the income distribution data I had available was broken down in this way. The analysis proceeded by using time series es·t.imates of the demand response to increases in income for estimating cross-sectional effects. They are no·t. the same things. Normally the cross-sectional effects would be higher and using time series estimates as stand-ins would overestimate consumption of the poor. El-Issaway( 1982) computes tha·t the lowest 40% of the population has 19% of total expendi·t.ure, based on data from 1974. This numb~r seems to have been fairly constant, since 1958, and I assume it will continue to be so. For 1990, if expendi tu.re continues to grow a·t i·ts historic ra·te, average expenditure will be 145 1977 L.E., so the poor will average 69 L. E. . Regretably, when these numbers were used wit,h the predicted 1990 prices the result was negative quantities of consumption for fed and veal meat. This is unfortunate firs·t of all because it suggests that when used for predicting future aggregate consumption demand may be shifted outward too rapidly by increasing incomes, and segondly because i·t prevents me from presenting any intelligent dis~ussion of meat consumption across .income groups. Cross sectional 50 demand for meat is obviously an area where further work would be very rewarding. 2000 Runs I made several runs for 2000, under various estimates of future herd sizes. The results paralleled those for 1990. The most distinguishing feature was dramatic price increases for meats. As there were no conclusions to be drawn from these results that could not be found in the 1990 runs I do not elaborate on 2000. Conclusion In terms of the evaluation of current and fu·ture policies, I conclude that present policies in the meat, sector cause substantial reductions in welfare, on the order of 3% of GNP. Alternative import orien·t.ed ·policies would provide more meat at lower cost. For the fut~re, imports on a large scale are necessary if meat consumption is to increase. Further research in this area would probably not be useful without improvements in the quality and ·t.ypes of data available .. Accurate values for herd numbers and slaughter quanti ties are essential, as is be·tter data on input-output 51 coefficients at the farm level. Cross sectional da·ta on meat consumption, by type of meat, would be useful. Specific information on the meat rationing program and its effectiveness would be helpful in planning alternative approaches to mitigating the effects of in~reasing price~ on the poor. 52 REFERENCES CITED 53 REFERENCES CITED Cuddihy, W.: Agricuiturai Price Management in Egypt. World Bank Staff Working Paper No. 366, The World Bank, Washington D.C., 1980. Dorfman, R., .P. Samuelson, and R. Solow: Linear Programming and Economic Analysis. McGraw-Hill, New York, 1~53. Duloy, R. . and P. Norton: "Prices and Incomes in Linea·r Programming Models." American Journal of Agricultural Economics. pp. 116-120, November, 1975. Enke, S. : "Equilibrium among Spa·t:.ially Separated Markets: Solution by Electric Analogue." Econometrica. pp. 40-47, January, 1951. FAO: The Water Buffalo. FAO Animal Production and Healthi Series, No. 4 .. FAO, Rome, 1977 Fitch, J., and I. Soliman: The Livestock Economy in Egypt: An Appraisal of the Current. Si tuat.ion. Economics Working Paper Series, No. 29, Agricultural Development Systems: Egypt Project,- University of California, Davis, 1981. Ikram, K.: EgyEt: Economic Pevelopment in a Period of Transition. Johns Hopkins University Press, Baltimore, Maryland, 1980. Kearl, L., et alia: Arab and Middle East Tables of Feed Composition. Utah Agricultural Experiment Station, Logan Utah, 1983. Kennedy, P.: A Guide to Econometrics. MIT Press, Cambridge Massachusetts, 1979. Mariano, R.: "Analytical Small-Sample Distribution Theory in Econometrics: The Simultaneous-Equations Case." International Economic Review. pp. 50~-533, October, 1982. 54 Martin, N.: "Stepped Produc·t. Demand and Factor Supply Functions in Linear Programming Analyses." American Journal of Agricultural Economics. pp. 116-120, February, 1975 Mohie-Eldin, A.l "The Development of the Share of Agricultural Wage Labor in the National Income of Egypt." in The Political Economy of Income Dis·tribution in Egypt, G. Abdel-Khalek and R. 11'ignor eds .. Holmes and Meier, New York, 1982. Murtagh, B., and M. Saunders: MINOS: Users Guide. Technical Report SOL 79-100, Stanford Universit~, 1979. Norton, R. arid Leopoldo.M. eds.: The Book of CHAC. Johns Hopkins University Press, Baltimore, Maryland, 1983. National Research Council: Nutrient Requirements of Beef Cattle. sixth revised edition, National Academy Press, Washington D.C., 1984. a. Preston, R., McConnen, and G. Haynes: An Analysis of Red Meat Production in Egypt. Unpublished Report, International Agricultural Development Service and Ministry of Agriculture, Egypt. 1984 Richards, A.: Egypt's Agricultural DeveloEment, 1800-1980. Westview Press, Boulder, Colorado, 1982. Samuelson, P.: "Spatial Price Equilibrium and Linear Programming." American Economic Review. pp. 283-303, June, 1952 Shapouri,· S., and I. Soliman: Egyptian Meat Market: Policy Issues in Trade, Prices and ExEected Market Performance. ERS Staff Report No. AGES841217,. USDA, Washington D.C., February 1985. Soliman, I.: Red-Meat Price Policy in EglE1· Economics Working Paper Series, No. 62, Agricultural Development Systems: Egypt Project, University of California, Davis, 1982. . Soliman, I., and M. El-Azim: An Appraisal of Lives·tock Concentrated Feed Policy Ill Eg~. Economic~ Working Paper Series, .No. 138, Agricultural Development Systems: Egypt Project, University of California, Davis, 1981. 55 Soliman, I., T. El-Zaher, and J. Fitch: Hilk Production Systems in Egypt and the Impact of Government Policies. Economics Working Paper Series, No. 121, Agricultural Development Systems: Egypt Project, University of California, Davis, 1981. Soliman, I., J. Fitch, and N. El Aziz: The Role of Livestock Production on the Eg:ypt,iap. Farm. Economics Working Paper Series, No. 85, Agricultural Development Systems: Egypt Project, University of California, Davis, 1982. Winrock International: Po·tential for On-Farm Feed Production and Utilization by the Egyptian Small Farm Sector. Winrock International, Morrilton, Arkansas, 1980. . . World Bank: 1985 World Development Report. Johns Hopkins University Press, Balt,imore, Maryland, 1980. 56 APPENDICES '. 57 APPENDIX A Crop and Nutrient Coefficients A portion of the model is an abbreviated description of the Egyptian crop sector as it applies to lives·tock. Activities with negative objec·tive function values convert labor and land into crops and crop by-products. Activities exist for selling the crops and transfering the relev,ant by-products to the livestock sector where they are available for feeding to animals. Further activities transfer to the various animals in various quarters, converting t.he feeds into nutrients in the process. This appendix will explain the derivation of the objective function coefficients for the cropping activities, the labor requirements, land availability, ·coefficients for the amounts of crops and feeds produced, selling prices for the crops, the seasonal pattern of feed availability, and the nutrients available in the feeds. Objective Function Coefficients Several sources were available with information on the variable nonlabor costs of planting for the base years, 58 1977-1978. Winrock(1980), Ikram(1980), Richards(1982) were all examined. Unfortunately these sources gave grossly different estimates for the costs, and none made any distinction betMeen farm sizes. Given this situation and the relatively unimportant effec·t of these costs on the cropping pattern allowed by the model, I chose the simple expedien·t of taking the average of all the available es·timates. Labor Requirements Labor requirements for crop produc·tion were derived from data in Richards and in the Winrock study. Winrock (pp. 113, 126) gives labor requirements for crops by mon·ths for each study village, while Richards (p. 233) gives three estimates of labor requirements, for the labor of men and women and children. I decided that Richards' dat.a was the mos·t reliable, largely on the basis of its comple·teness. Since the Winrock data offered a breakdown of requirements by months, I used their numbers to derive requirements needed each mon·th. percei~tages of total I then combined Richards' middle estimates of Male and Female labor required, accepting Mohie-Eldin's (1982) argument that since female wage rates were half men's their productivity was also half. Finally I divided up these total requirements using the monthly percentages calculated from Winrock. For Berseem, I 59 followed basically the same procedure, making some assumptions about how requirements for the berseem varied for different types. The following table shows the resul·ts. TABLE 6. Monthly Labor Requirements. Man Days per Feddan. Cotton Rice Maize Sorghum Wheat Berseeml J F M A M J J A s 0 N D 10.2 17.0 1.4 10.6 17.7 8.9 9.5 6.4 3.4 2.6 2.0 14.9 8.8 12.7 6.8 3.0 5.5 6. 7' 6.0 4.5 12.7 5.1 4.5 7.6 7.6 1.3 21.0 1.0 1.5 1.0 1.0 1.0 19.0 5.5 0.5 4.0 2.0 2 3 4 7.0 7.0 7.0 7.0 7.0 7.0 7.0 7.0 15.0 7.0 7.0 3.0 1.0 3.0 1.0 3.0 3.0 1.0 1.0 Land Availability For the base years of the study, I used as overall land availability the 1975 - 1978 average of crop area planted. (Ikram(1980), p. 414) Different totals are available for the summer and the ~inter plantings. For the calibration runs, used the 75-78 averages of area planted to each crop as additional constraints. Berseem was divided into 4 types, following Richards. Future years use these same coefficients, most of the literature says that land development schemes are barely, if at all, keeping up with losses from urbanization. The land areas had to be divided I 60 between small and large farms, for which I used the figure that 50.62 percent of the crop land was held by small farmers. (Calculated from M~hie-Eldin(1982), p. 277.) In calcula·t.ing this I assumed that the land repor·t.ed as being used for orchards was all held by large. farmers. Again, for the ca.libration runs this ra·tio divided the area plan·ted to each crop proportionately between small and large. It seems unlikely that this is ac:;:tually true and there is anecdotal evidence that different sized farms plarrt different crops, but no usable numbers. For subsequent model runs, areas planted were allowed to fluctuate, see the chapter on results for details. The numbers used follow, in thousands of feddans. TABLE 7. Cropping Areas. Thousands of Feddans. Crop Total Small Farms Large Farms Summer: Cotton Rice Maize Sorghum s. remainder 5821 1302 1051 1846·· 451 1171 2947 659 532 934 228 593 2874 643 519 912 222 578 Winter: Wheat T. Berseem Berseem 1 Berseem 2 Berseem 3 Berseem 4 remainder 5228 1344 2801 844 383 183 1391 1083 2646 680 1418 427 194 93 704 548 2581 664 1383 417 189 90 687 w. 5~5 ---------------------------------------------------- 61 Crop and Feed Production Coefficients The sources for these coefficients are Winrock(1980), Soliman and El-Azim(1981), and Ikram. Where there was a discrepancy between the sources over the same coefficient, generally the case, the average was used. TABLE 8. Crop and Feed Coefficients. Tons per Feddan Crop Coefficients Cotton Rice Maize Sorghum Wheat Fodder: Berseem Berseem Berseem Berseem .919 2.212 1.583 1. 591 1. 417 1 2 3 4 Lint Grain Grain Grain Grain .39 Seeds 1.2 S·t.raw . 2 Leaves 1.485 Straw .184 Bran .153 Bran 6.5 13.0 19.5 22.0 Selling Prices Selling prices were no problem for the base runs of the model. Prices for crops are set by the government. For later runs assumptions had to be made. One important one was that the government would relax the mandatory area requirements. Specific details are covered in the chapter on results. For the base runs, selling prices for grains and cotton were as follows. The source is the Ministry of Agriculture, cited in Ikram, p. 424. 62 TABLE 9. Selling Prices for Crops. L.E. per Ton Crop Price Wheat Maize Rice Cotton Sorghum 61.7 71.4 66.1 222.0 . 65.0 Nutrients from Feeds Data on the nutrients. in feeds comes solely from Kearl(1983). They were recalculated so as to be equivalent to the units used in the NRC formulae. The results of this were as follows. TABLE 10. Feed Nutrients. Grams per Kilogram. Crop Nutrient CP NEM NEG 920 292" 1410 863 930 870 33 105 524 1106 35 610 880 270 96 19 1704 370 1137 213 890 170 96 28 1670 227 1103 129 930 890 890 35 144 93 863 1364 1777 357 835 1196 850 190 1373 858 DM Cotton seeds Rice: straw bran Maize: Grain leaves Sorghum: grain Berseem fodder Wheat: straw bran Imported Maize Unified Feed (1977 formula) ------------------------------------------------------- 63 APPENDIX B Livestock Coefficients This appendix explains the deriva·t:.ion of the nutrient requirements, and the parameters that describe herd performance, composition and milk production. Requirements The nutrient requirements for livestock were derived using the functions in NRC(1984). There are several problems with this procedure. First of all they are obviously not correct for water buffalo, which is a different genus. They were used· because no other numbers were available. Secondly, even for cattle they are subject to critic~sm. The major complaint is that they allow no substitution between nutrients, when such substitution certainly occurs in fact. Again, no acceptable alternatives are available. Requirements for dry matter intake (MNDM), crude protein (CP), net energy for maintenance and for gain (NEM, NEG), and maximum dry matter intake (MXDM) were calculated. The formulae used are as follows. ' 64 MNDM For breeding females, MNDM = W**.75 * (.1462*NEM- .0517*NEM**2- .0074). For growing and finishing cattle, MNDM = W**.75 * (.1493*NEM- .0460*NEM**2 - .0196), where W equals live weight, and NEM equals Meal NEM per kg. diet. NEM was calculated as a weigh·ted average of the NEM in the . available feeds. This varies with the season. Adjus·tments ·to this figure were made on the basis of sex and frame size. CP CP = (33.44 * (MNDM+MXDM)/2 + 2.75 * W**.5 + .2 * W**.6) I (. 90 * . 66), with additional requiremen·ts of 55 g/day for the last 3 months of pregnancy and of 33.5 times milk production in kilograms during lactation. NEM NEM =C * W**.75, where C is a temperature dependent constant equal to .084 during quarters 1 and 3, .077 during quarter 2, and . 075 during quarter 4. Addi·tional NEM is required for pregnancy and lactation, calculated from NEl1 and = CBW * (0.0149 - .0000407t) *e**(.05883t-.0000804t**2) NEM = .1 * ( %fat ) + .35 where t is the day of gestation, CBW is calf birth weight. 65 NEG NEG = .0608 * (W **.75) * (ebg)**1.119, where ebg is empty body weight gain. MXDM MXDM = .035 * W These formulae were used in a computer program to estimate nutrient requirements for the model. The program set the variable parameters to their proper values for the various types of animals, calculated basic requirements on a daily basis, added in pregnancy and lactation requirements, added in weight gain where appropriate, and then accumulated the quarterly totals. Herd Parameters Literature The data used in deriving the parameters of the livestock herd comes from 2 main sources. The first source is Winrock(1980). This study included livestock surveys of two villages, one in the Nile delta and one in upper Egypt. The data collected included the number and type of animals held by different sized faims, death and birth rates and the month of birth, weights at birth and ma·turity, culling age, 66 milk production, labor inpu·ts, and so on. In addition to this survey, data from several working papers published by the Agricultural Development Systems (ADS) Egypt Project were used. The data in these papers comes from the 1977 Farm Management Survey. Herd Structure Average data from the two Winroclt surveys p:t."ovided the starting point for the herd structure coefficients. '!'here are two main coefficients, one representing the number of calves born yearly from each breeding member of the herd, and another representing the number of animals needed each year to replace cull and death losses and provide for growth in the herd size. The calving coefficients were calculated from the data on parturition inteJ.~vals and birth ra·tes. Again using cattle as an example, the annual birth rate was 63.2% and the parturition interval 15 months. Thus 79% of the breeding females conceive and hav~ a live birth every 15 month cycle. Assuming a 5% abortion rate, 83% in the 15 month cycle or 66% annually, conceive, and so .63 live births occur per year. 17% of ~he breeding females do not concieve and so require only nutrients for maintenance for a year. (After a 12 month period I assume they conceive or are culled.) In addition to these births, the replacement 67 breeders produce a calf as a heifer before they enter the breeding herd. Fo.r.: buffalo the corresponding numbers are 55.3% annual birth .r.:ate, an 18 month par-turition interval, 83% with live births each 18 months, 87% concieve each cycle, 13% on maintenance. Death Losses These coefficients also come from the Winrock s~rveys. Data were reported on calf and adul·t. mortality ra·t.es. These rates were used to calculate both expected losses for breeder replacements and feeders and the nutrients consumed by a typical. animal expected to die. For example, the typical Baladi steer ra~sed as a feeder has a mortality rate of 15% in the first six months, 3% in the second, and 4% in latter years. Since I have assumed it takes two years to raise a feeder to the point where it goes ~o the feedlot, the survival rate for this type animal was calcula·ted as ,. . 79. Every unit of the feeder activity would then require .21 of the death loss activit,y. This activity takes an animal out of the calf balance, and also takes nutrients out of the nutrient balances. These nutrient requiremen·ts were calculated as a fraction of those required to raise an .' ' 68 a~ animal a£ this type to maturity. I£, for example, 80% ' those animals that die do so in the firs·t quar·ter, nu·trien·t requirements in the second quarter will be 20% a£ those required for a live animal. For cattle raised as breeders, for 3.5 years, the death loss requirement is 26%. For buffalo Winrock's loss numbers are 21% the first 6 months, 4% the second, and 4% for latter years. The .loss coefficients are 27% for feeders and 31% for breeders. The 4% loss rates also apply to ·the breeding animals, ·thus to maintain them a certain number a£ replacements are needed annually. Winrock repbrted that the normal culling ages were 9 for ,cattle and 11 for buffalo. I calcula·ted ·the proportion of relacements needed by calculating losses over the time in the herd and then adding the loss of t,he animals at culling age. The results were that to maintain a constant size herd 18% of the cattle and 14% o£ the buffalo had to be replaced anually. In reality the herds have been growing in recent years. F~tch(1981) suggests a rate of 2.5% for cattle and 3% for buffalo, although the data are poor. Since each breeding female in the cat·tle herd produces 4. 27 calves over her lifetime, or .776 per year, this suggests the breeding herd (HERD) is growing at 2.5 / .776 = 3.2%. For buffalo the number is 4.5%. These numbers were added to the above replacement requirements to get .22 and .19, the proportions 69 of the cattle and buffalo herds which must be new breeders each year. Growth Rates This was one of the less scientific parts of the procedure. Data on birth and cull weigh·ts was available f l."om Winrock, along with some sketchy information on the weigh·ts of feeder animals, from Soliman(1982) In addition Winrock gave average gain rates for immature animals. The Winrock study also reported that animals were not fed sufficient feed for weight gain except during the six month season (winter) when berseem was available. From this information tables were constructed that allowed animals to reach their mature weight at the specified age, while not growing during the summer, exc,ept in the case of the young. The grow·th rates used assumed that gro~th would occur at an initially increasing, then decreasing rate. Milk Milk Production Coefficients Data on milk comes from the Winrock study, the 1977 Farm Management Survey, FA0(1977.), and NRC(1980). 70 Winrock data gives average milk production as 1080 kgs. of 7% fat for Buffalo and 666 kgs. of 3.5% fat for cattle, per lactation. This was assumed to be spread over a seven month lactation period, with the 2nd ·through 4th months having production roughly 1. 7 times that of the o·ther four months. If calves are kept, versus sold for veal, a certain amoun·t of the cows milk is consumed by the calf. There was not data on what these quantities typically are, or on Egy;ptian weaning ages. I made the arbitrary assump·tion,s '·that all the first months milk goes to the calf, along with one kg. a day for the subsequent six months. The result was as follows. TABLE 12. Milk Coefficient.s. Kilograms. Period Birth month + 1 quarter + 2 quarters Production 72 378 216 Cat·tle Calf req. Production Buffalo Calf req. 120 600 360 120 90 90 72 90 90 In the model, the calving activities add the production less calf requirements to the balances, while sell{ng the calf for veal adds the calf requirement back. Milk Prices These vary according to farm size, the averages for my two farm sizes from the FMS are reported here. 71 TABLE 13. Dairy Product Prices. L.E. per Kilogram.· Product Raw Milk Ghee Raw Cattle Milk Raw Buffalo Milk Skim Milk Farm Size Small Large .154 1.264 .14 .15 .12 .143 1. 452 .15 .16 .13 Buffalo milk has a higher fat percen·tage and is worth more. Anecdotal evidence in Soliman, El-Zaher, and Fitch(1983) was that small farms recieved lower prices than large. As a result of these facts I rather arbitrarily made up some prices for the model, as shown above in Table 12. Prices in reality will vary greatly with distance from cities, this was ignored. Ghee Production Coefficients The FMS provided substantial da·ta on this subject, though some assumptions about the yields of different types of milk had to be made. These were based on fat percentages. Cattle Milk: 1 kilo milk yields .15kg cream and .85 skim milk. The cream then gives .066kg butter and .084 buttermilk. The butter produces .042kg ghee. Buffalo Milk: 1 kilo milk yields .15kg cream and .85 skim milk. The cream then gives .132kg butter and .018 buttermilk. The butter produces .084kg ghee. 72 Milk Labor Requirements I guessed that the equivalent o£ 3 days per month were spent milking d~ring lactation. Producing ghee from milk took .05 days per kg. of milk. 73 APPENDIX C Labor Coefficients Labor is supplied to the model in several forms. First, certain quantities of male and female (I included children in this category) laboz: come from the farm fami.lies. These supplies are assumed to be fixed. Additional .quanti·ties of male labor can be hired, at rates which vary according ·to the season. As well, male and female labor can be transfered back and forth. This appendix will attempt to explain the derivation of the quantities of available labor, ~he prices for hiring labor, and the coefficients which determine the rates at which different kinds of labor act as substitutes for each other. Quantities of Labor on the Farms Several attempts to get reasonable estim~tes for these numbers were made, but it was not to be. Mohie-Eldin(1982), reports data from various farm surveys, but includes stern caveats on their use. I made the simple assumption that one male laborer and 2 female/child laborers were available for each farm, and that these workers could each provide 20 days per month for the types of crop and livestock labor tha·t are 74 included in the model. These quantities are rather low, but the model includes only part of the actual labor required to run a farm. These quanti ties were mul·t.iplied by the number of farms in each size category, and the results were used as right hand side values in the model. Wage~ for Hired Labor This data comes from MOA data cited in Mohie-Eldin, p. 254. TABLE 13. Male Labor Costs. L.E. per Day Month J F M A M J J A s 0 N D Wage .76 .77 .80 .83 .84 .88 .92 .92 .92 .92 .93 .93 Transfer Coefficients for Labor This also comes from Mohie-Eldin. He argues, based on the wage difference, that female labor is half as productive for field work as male. I have assumed in addition that when 75 male labor is used for such tasks as tending livestock, milking, and producing ghee it is no more productive than female. 76 APPENDIX D Computer Program The results in this thesis were calculated using MINOS, a linear and nonlinear minimization program for problems with linear constraints. The program was developed for the Navy by Bruce Murtagh and Michael Saunders of the Stanford Department of Operations Research. :MINOS uses a version of the revised simplex method for linear problems. For nonlinear problems, additional "superbasic" variables can be used, which vary to improve the objective function while the basic variables change to keep the solution feasible. For problems with few nonlinear varibles, such as mine, the quasi-Newton method is used to pick the direction and size of the change in the superbasic variables. Details are available in MINOS~ N0nlinear Programming System (For Problems A Large Scale ~ith Linear Constraints) User's Guide, report SOL 77-9, Department of Operations Research, Stanford University, Stanford California. l