Determining the optimum investment pattern in information technology on dairy farms

Transcription

1 Van Asseldonk et al.: Determining the optimum investment pattern in IT 81 Determining the optimum investment pattern in information technology on dairy farms 1, M.A.P.M. van Asseldonk, R.B.M. Huirne, A.A. Dijkhuizen, A.J.M. Beulens 1 Department of Farm Management, Wageningen Agricultural University, The Netherlands 2 Department of Computer Science, Wageningen Agricultural University, The Netherlands Abstract Benefits, as a result of improvement in biological parameters and reduction in labour costs, were evaluated for information technology applications that support operational management on dairy farms. The applications evaluated included pedometers to measure daily physical activity of cows, automated concentrate feeding systems and on-line automated parlour systems for recording of milk production, milk temperature and electrical conductivity of quarter milk. The benefits were accounted with respect to concentrate feeding, oestrus detection and mastitis detection and were based upon modeling techniques taking into account farm-specific characteristics and accuracy of the system. The accuracy of the systems were elicited from expert opinions using conjoint analysis. In the next phase of the project, investment decisions will be examined using a stochastic dynamic programming framework enabling to take into account the fact that farmers can adopt improved technologic components in a stepwise manner. In addition, improvement in accuracy of the systems and trends in investment costs over time can be incorporated. In this way the investment policy that maximizes the present value of expected net returns of a strategic plan for an individual farm over a given time horizon can be determined and quantified. 1. Introduction Recent advances in information technology (IT), such as on-line data of pedometers (PM), milk production measurement (MPM), electrical conductivity of quarter milk (ECQM), automated concentrate feeder (ACF) and milk temperature measurement (MTM), have increased the opportunities for effective management support. At the present time a multi-factorial implementation is becoming feasible in which several IT applications are implemented at the same time to support concentrate feeding, oestrus detection and mastitis detection. In 1993, several IT applications were already implemented on Dutch dairy farms (number of farms between brackets), such as pedometers to measure daily physical activity of cows (200) and automated concentrate feeding systems (12000). Moreover, on-line automated parlour systems were adopted for recording of milk production (1700), milk temperature and electrical conductivity of milk (70) (ATC, 1994). In most cases different stepwise manners (scenarios) of IT implementation are possible. However, the effects of an individual IT implementation and the different scenarios on the technical and economical results are yet unclear. Up to now, the

2 82 Dutch/Danish symposium on Animal Health and Management Economics, 1997 commercial viability of the IT applications and combinations has not been undertaken. An economic evaluation of IT applications and possible IT scenarios can be based upon normative as well as positive approaches (Verstegen et al., 1995). Field data for a positive approach are very limited however, and previous empirical research mainly documented the impacts of a specific combination of IT applications (Carmi, 1992; Gelb, 1996). Furthermore, the approaches reported did not decompose the overall impact of the multi-factorial approach to the individual business functions/processes. The necessary experiments to determine the individual and combined effects of the multi-factorial approaches would be costly and they would take years to complete. Moreover, the forecasting of future benefits of the possible scenarios of IT implementation are difficult to determine with historical farm data due to the dynamic development in sensor research and development, and analysis techniques, outdating previous research. In this research a normative approach is used in order to support investment decisions. The objective of this paper is to discuss the method of deriving economic benefits of pedometers, concentrate feeders and measurement of milk production, milk electrical conductivity and milk temperature. In the near future, results will be used to model the impacts of different IT applications and scenarios in order to support investment decisions. The necessary input will be discussed with respect to detection characteristics and potential benefits, followed by an outline of the dynamic programming model. 2. Quantifying detection characteristics of information technology In order to derive the potential economic benefits, detection characteristics of different multifactorial IT implementations are necessary. However, the detection characteristics in most experimental studies are not decomposed to the individual effects in the multi-factorial approach and are obtained in experimental research stations which do not necessarily reflect the possible impacts on a commercial farm. If actual data are missing and therefore analysis of objective data is not possible, necessary information can be obtained from expert opinions to interpret the implications of reserach results for commercial farms, resulting in subjective analysis. The method of conjoint analysis is used to elicit the opinions of experts. The method of conjoint analysis was developed from the theoretical work of mathematical psychologists (Luce and Tukey, 1993) and is commonly used in marketing research (Cattin and Wittink, 1982) for measuring buyers= tradeoffs among multi-attributed products and services. The tradeoff or importance of each attribute is assessed indirectly by decomposing the overall preferences for several multi-attributed alternatives. Main advantage of the conjoint method is that the predictive performance outperforms the compositional method in which the respondents have to assess the values for the attribute levels directly (Huber et al., 1993). In this study conjoint measurement is used to quantify and predict the experts= overall judgement of a particular multifactorial detection system on the basis of the underlying applications (called "attributes"). The experts= overall judgements for a set of predefined alternative detection systems (called "profiles") are broken down in order to determine the contribution (called "part-worths") of each application within the detection system (Green and Srinivasan, 1978). The two basic assumptions are that 1) the detection system could be described by a set of applications and 2) the overall judgement of the experts about that system were based on the underlying applications. The described applications were used as independent dummy variables in the conjoint model with

3 Van Asseldonk et al.: Determining the optimum investment pattern in IT 83 a value of 1 if the IT application was implemented and a value of 0 if the application was not implemented (see Equation). SE OD '61%15(PM%5(MPM%1(ACF%5(MTM&3((PM(MPM)&2((PM(MTM) SP OD '87%1(PM%1(MPM%1(MTM E CMD '10%5(PM%16(MPM%44(ECQM%2(ACF%16(MTM&8((MPM(ECQM)&5((MPM(MTM)&7((ECQM(MTM SP CMD '71%1(PM%4(MPM%8(ECQM%1(ACF%1(MTM&1((MPM(ECQM)%1((MPM(MTM)%1((ECQM(MTM) SE SP OD CMD PM / MPM / ECQM / ACF / MTM = sensitivity; = specificity; = oestrus detection; = clinical mastitis detection; = attribute dummies. From the models, detection characteristics of the individual IT applications can be derived by adding the corresponding part-worth of the main effect to the constant term. A part-worth of the first order interaction must be included if a specific combination of IT applications are of interest. For example, a complete multi-factorial implementation to detect oestrus results in a sensitivity of 82% ( ). 3. Potential economic benefits of information technology Simulation techniques can be appropriate methods to evaluate IT applications that are related to operational functions effecting labour costs and benefits resulting from an improvement in biological parameters. Simulating a specific farm, with and without the IT application, provides a method to determine the impacts of the improved biological parameters (Verstegen et al., 1995) taking into account current status of farm performance (e.g. oestrus detection, average milk production per cow per year, mastitis incidence, roughage quality and feeding strategy), farm structure (e.g. farm size) and accuracy of the system. Since the exact improvements are unclear, simulation models can be applied for sensitivity analysis, enabling to gain insight into the possible benefits and to perform break-even analysis (Kleijnen, 1984). In order to obtain benefits as a result of the different improved processes, and ultimately to determine the effects of the different investment alternatives, the analysis should include oestrus detection in the function reproduction, mastitis detection in the function health care and feeding in the function nutrition. Oestrus detection can be supported with a number of IT applications, an increased activity and milk temperature as well as a decreased electrical conductivity, milk production and concentrate feed intake are associated reactions of oestrus. An increased electrical conductivity and milk temperature and a decreased activity, milk production and concentrate intake are signs of mastitis. The expected effects of an improved management due to an early detection of potential clinical mastitis cases could be a reduction of clinical mastitis incidence, a better responses rate to treatment, less lactation therapy with antibiotics, less production loss due to clinical cases, less premature culling and a reduction in cow and/or bulk milk tank somatic cell

4 84 Dutch/Danish symposium on Animal Health and Management Economics, 1997 count. However the effects of the different therapy systems and strategies remains to be shown.implementing an automated concentrate feeder and an on-line milk production measurement accomplishes an automated yield orientated feeding in which the supplementation of concentrates is optimized. An automated concentrate feeder provides a method of norm feeding in which concentrates are supplemented to individual cows and a daily milk production measurement provides an increased accuracy of the norm feeding level in comparison with a 3 or 4 weeks measurement. The TACT simulation model, described by Jalvingh et al. (1993), is used for the function reproduction and some modifications were applied to include effects in the function nutrition. The optimization model of Houben et al. (1994) was used to simulate the effects of clinical mastitis detection. Identical basic prices and other parameters were used in the different simulation models 5. Discussion and outlook In this study the method to determine the potential economic benefits of the five mentioned IT applications is described. Investment and operation costs as well as the detection characteristics for the different possible multi-factorial approaches should be added to the assessed economic benefits. Since the detection improvements and their effects on biological parameters are unclear, the simulation models can be applied for sensitivity analysis, enabling to perform an economic evaluation and break-even analysis for the IT applications and IT combinations. A break-even analysis is particularly interesting in case of applications that support mastitis detection. Up to now, the effects of therapy of early diagnosed clinical mastitis quarters are unclear and the breakanalysis will quantify the minimum effects necessary to justify the investment. The investment decisions will be examined using a stochastic dynamic programming framework. The model's objective is to determine the investment policy that maximizes the present value of expected net returns over a given planning horizon. At the beginning of the planning horizon, the model starts with a farm without the five IT applications. The model will determine and quantify the optimal combination of a stepwise implementation of applications for an individual farm and will be able to include uncertainty on the input-output relationship. The optimal strategy is determined under different farm structures with respect to farm size, milk production per cow, feeding regime, level of oestrus detection and mastitis incidence. Inclusion of the years in operation of the milking parlour is important because a combined renovation of the milking parlour and investment in parlour automation is cost reducing. References ATC, Automizing in the agriculture sector: use and trends (in Dutch). Wageningen, The Netherlands. Carmi, S., The performance of an automated dairy management data-gathering system. In: Proceedings of the international symposium on prospects for automatic milking, Wageningen, The Netherlands, November 23-25, EAAP publication 65: Cattin, P.; Wittink, D.R., Commercial use of conjoint analysis: a survey. Journal of Marketing, 46: Gelb, E.M., The economic value of information in an information. In: Proceedings of the 6th international congress for computer technology in agriculture (ICCTA), June 16-19, Wageningen, The Netherlands;

5 Van Asseldonk et al.: Determining the optimum investment pattern in IT 85 Green, P.E.; Srinivasan, V., Conjoint analysis in consumer research: issues and outlook. Journal of Consumer Research, 5: Houben, E.H.P., Huirne, R.B.M., Dijkhuizen, A.A. and Kristensen, A.R., Optimal replacement of mastitic cows determined by a hierarchic Markov process. Journal of Dairy Science, 77; Huber, J.; Wittink, D.R.; Fiedler, J.H.; Miller, R., The effectiveness of alternative elicitation procedures in predicting choice. Journal of Marketing Research, 30: Jalvingh, A.W., Van Arendonk, J.A.M. and Dijkhuizen, A.A., Dynamic probabilistic simulation of dairy herd management practices. I. Model description and outcome of different seasonal calving patterns. Livestock Production Science, 35: Kleijnen, J.P.C., Quantifying the benefits of information systems. European Journal of Operational Research, 15: Luce, R.D.; Tukey, J.W., Simultaneous conjoint measurement: A new type of fundamental measurement. Journal of Mathematical Psychology, 1: Verstegen, J.A.A.M., Huirne, R.B.M., Dijkhuizen, A.A. and Kleijnen, J.P.C., Economic value of management information systems in agriculture: a review of evaluation approaches. Computers and Electronics in Agriculture, 13:

6 86 Dutch/Danish symposium on Animal Health and Management Economics, 1997