In 2007 MAF policy engaged Four Cubed Limited to explore the potential from optimising resource allocation on NZ dairy farms. The assignment was the outcome of discussions over the preceding year with Treasury and MAF Policy by Four Cubed and others on subjects that could broadly be described as agricultural production economics. It was expected that the results would be published, or at least presented in an appropriate forum.
MAF circulated a draft of the results around interested government agencies and industry parties. The dairy industry hotly contested aspects of the work. No attempt was made to formally present or publish the findings, but agprodecon.org received a boost.
The passage of time reveals much. Optimising resource allocation is now entering the lexicon of the dairy industry, the 2009 Dairy Industry Strategy’s focus is on systems – systems of course being the basis of production economics, and Agresearch and Lincoln University have been jointly funded $1 million to develop LP farm models. The latter should be questioned given that the technology already exists, is NZ owned and developed, works well and is available commercially.
The following is work done for MAF Policy as it should have been published in August 2007.
The findings are still relevant today, but would be more precise if redone with current costs. A rough heuristic given increases in on-farm costs would be that farm operating surplus in 2006 from a milk solids payout of $4.24 is equivalent to that from a $5.20 payout in 2009.
The authors share a view that the perceived linear relationship between production and profit for NZ dairy production - which had some basis through the 20th Century - no longer holds.
In part, simple overproduction (allocative inefficiency) means that lower levels of production would provide dairy farms higher levels of income. Nationally, gains of $252 million in operating surplus would have accrued from dairy farms producing where MC=MR in 2006 – an average of 23 cents per Kg of milk solids or a gain in operating surplus of 24% on a per farm basis.
These numbers are based on regional average farms. Averages tell us little about the distribution of performance within the sample. The numbers though are consistent with anecdotal evidence and commercial experience. They have a possible bias to understatement.
A range of improved farm management options will additionally provide opportunities to improve operating surpluses through productive efficiencies – some also resulting in lower resource use and output.
Many resource inputs and management practices are highly interchangeable and may produce little difference in operating surplus. Some are highly sensitive to costs.
This work has begun to clarify the information possible from simulation modeling, and the form and presentation of such information required to make more efficient investment or resource allocation decisions.
More extensive use would be expected of these tools in farm management if modeling continues to clarify the information required for improved farm decision making, and the presentation of that information gains a wider audience.
The most important factor in improving dairy farm operating surplus however remains in setting the correct production level relative to the cost and availability of resources for the prevailing payout. Given a) the scale of additional operating surplus available to dairy farms from setting the correct production level and b) the ready availability of software capable of accurately providing such analysis at reasonable cost, it would be reasonable to rely dairy industry leadership to fully support such initiatives.
That would leave the major question being how best to communicate information on optimal dairy farm production levels. This will vary depending on the economic literacy of the audience. Suggestions are:
The project’s primary objective was to determine whether there are significant differences between existing farm resource use and profitability, and those suggested by techniques optimising resource allocation. That being the case, to then provide indications of the economic gains in terms of changes in farm profitability, improved resource use and any reductions in societal costs such as emissions from the application of such techniques on New Zealand dairy farms.
This would also allow MAF Policy to confirm the veracity of simulations from bio-economic models and the GSL model in particular.
It was expected the project would suggest possible applications for modeling improvements in allocation of agricultural resources including shadow prices for agriculture’s societal costs – nitrogen and water use, and emissions in particular.
The Grazing Systems Limited (GSL) Model is a Linear Program model for pastoral farm systems. It optimizes animal production needs against dry matter feeds (energy) – pasture, crops and supplements. It is a bio-economic model in that resources have economic values that drive optimization.
GSL is an infinitely recursive model with multiple subsystems. It normally operates a 26 period year but can be run 4 weekly. It is a closed system returning animals and pasture to their condition at the start of the year.
Core inputs to the model are normally a pasture profile and growth and production patterns. For pasture, GSL:
For animal production on dairy systems GSL works with user prescribed:
Output prices can be set independently for each output for each period.
Commercial use of GSL is normally an interaction between a farm advisor and the farm manager to establish pasture profiles, animal production objectives, expected output prices, business objectives, attitude to risk and management practices, and then setting constants and constraints. The model is left to maximize profit by determining optimal resource settings within these constraints. Ongoing adjustments should be made to resource settings as conditions and prices change or to establish a production/stocking rate “envelope” for various input functions.
When used as a modeling tool GSL generally takes an established model representing an optimized farm system and varies a single input parameter about the optimal – typically herd size but it can also be herd structure, calving dates, animal or pasture production, or management decisions on culling or drying off/sales. This provides data to build the system production function, and marginal cost and/or revenue curves.
GSL can be used to determine demand responses to prices of inputs such as supplements or nitrogen, or supply responses to output prices – most commonly milk solids payouts.
The ability to substitute use and management of specific resources as part of the optimisation process provides the difference between this and other more deterministic simulation methods.
The use of GSL is highly educational to those involved – it is not restricted to thinking inside the square and will often provide answers that are sensible but not necessarily intuitive.
In this project GSL was initially used in reverse to establish pasture and lactation profiles from MAF Regional Dairy Farm and other data. Farm accounting data provided average farm and herd size, cow milk production and costs for farm working expenses.
MAF provided the amount of nitrogen used in each region and the response rate of 10:1 (Kg dry matter (DM) utilized per Kg of nitrogen (N) applied - effectively about 13:1 Kg dry matter produced at modest pasture utilisation).
Some assumptions as to the quality, utilisation and price of supplements and the proportions used for each region were made based on research farm and/or monitor farm data.
Adjustments were made by overlaying some practical regional constraints for pasture utilization in winter (soils/rainfall) and known pasture quality differences (MJME/Kg DM) This along with the MAF data allowed the quantity, timing and quality of pasture – a pasture production profile – to be established for each region. These are averages of many farms and the timing of silage making and feeding, crops, nitrogen application and actual average pasture covers present may vary from specific farm practice.
Milk production pattern was modeled on Fonterra milk collection flows for the appropriate region.
Per cow costs were based on MAF data for average farm working expenses, but excluding those costs handled more precisely by the GSL model. Those costs are for nitrogen and its application, supplements made or bought and their feeding out, crops and grazing off including transport. It should be noted that per cow costs are not in reality flat but show a declining trend with lumpy increases as additional labour units or capital items are required with increasing herd size.
The GSL model was then used constrained (resource use specified rather than optimised) for each region to establish the base “average regional farm” model reflecting both farm capabilities and management.
It was not possible using regional average data to establish a model for the Canterbury region that would be both feasible and viable. This suggests that data for Canterbury derives from widely different systems producing a regional average that reflects neither. This is possibly due to the mix of irrigated and non-irrigated farms, and/or a strong presence of high production systems across the region.
The regional farm models created for Northland, Waikato, Lower North Island and Southland were then used to establish optimal regional solutions by removing the constraints on cow number and nitrogen use that had been required to model the MAF average farms’ performance. These models provide the optimal use of resources for the farm’s capabilities and their existing management system.
Each region’s average dairy farm results (standard and optimized ‘profit’) were extrapolated to a regional total using the regions percentage of the national milk production total and these then summed to a national total excluding the Canterbury region.
Curves representing marginal revenues per cow were constructed for Waikato and Southland about the optimal herd size at milk solids payouts of $4.50, $5.50 and $6.50. These curves clearly indicate the nature of the step change from marginal revenue positive to marginal revenue negative that occurs with pastoral farms about the optimal production level.
Optimal farm operating surplus performance for Waikato and Southland were modeled for milk solids payouts between $3.00 and $8.00 per Kg of milk solids – key data provided being herd size, operating surplus and milk production.
The economic impact of improving two management factors: a) improved pasture utilization (IPU) and b) better herd management (BHM) were investigated. These were done only for the Waikato MAF average model, and with the payout set at $5.50 per Kg of milk solids. The use of a $5.50 payout was in anticipation of a significantly improved payout for the 2007-2008 season.
Better herd management consisted only of changing the replacement policy of the model herd such that:
These three changes have the multiple synergistic effect of having a more mature herd in which proportionately more cows are producing at a higher level per cow. There is simply less young stock being reared (milk being consumed), eating grass or having to be grazed off at a cost.
Improved pasture utilisation involved scaling average pasture utilisation from slightly over 77% to 82%.
There is a belief in some quarters that increases in utilisation may lead to reductions in per cow performance. This could certainly be the case in farmlet trials where pastures are operated under ‘standard decision rules’, but in practice farms are able to perform as flexible systems where pasture management can vary both in grazing interval and grazing intensity. Differences in herd size and age structures allow grazing options without negative production effects.
Improved pasture utilisation and better herd management were also examined in combination and using a base pasture production level of 14.3 tons of dry matter per hectare reflecting a level of dry matter production expected of the Waikato region.
Regional profiles of the pasture dry matter energy available to cows and their lactation for the Waikato (Figure 1) and Southland (Figure 2):
The 89,380 Kg of milk solids production is the same level as that for the Typical Management level shown above in Table 2 for a milk solids’ payout of $5.50 per Kg i.e. the optimal production level for this farm with 2006 costs does not change between a $4.24 and $5.50 payout.
Allocative and Productive Efficiencies
There is more than one aspect to optimising farm resource allocation, and in that respect Figure 8 is a graph worth a thousand words to an agricultural production economist. The curves of operating surplus against herd size are in effect production functions. All display similar shapes representing increasing marginal costs as production intensity increases. The production level where operating surplus is maximized (marginal cost equals marginal revenue – MC=MR) is simple to determine. Setting production at the operating surplus maximizing point (MC=MR) on a production function is common wisdom and infers allocation efficiency.
That there are a wide range of possible production functions for the same farm should be no surprise given the complexity of pastoral farming and the diversity of farm managers and farm management systems.
What may be surprising to the agricultural production economist is how poorly the “Typical” farm production function stacks up against other available options, and the extent to which “Typical” farms produce beyond the point of allocative efficiency.
In NZ agricultural production economics is not widely practised, and agricultural production economists are few and far between. Before more detailed discussion it is worth stepping back a little to explain some production economic principles.
Management decides a production system that together with available resources determines the production function for that system e.g. decisions on choice of cow, calving date, herd management, pasture management and once or twice a day milking. Resource allocation decisions determine where on that production function the farm operates e.g. herd size, supplements fed, nitrogen used and drying off dates.
The operating surplus for the production system is maximized at the point where the marginal cost of an additional unit of production is equal to the marginal revenue from that unit. Marginal revenue and marginal cost are the respective revenue from and cost of producing that additional unit. The level of production where MC=MR may be higher or lower than the current level of production.
Determining the point where MC=MR requires knowledge of a production function as in Figure 8, a marginal revenue curve, or a marginal cost curve. A marginal cost curve may be less intuitive than the other two but has the advantage of being independent of revenue. Marginal revenue and marginal cost curves are in Appendices 1, 2, 3 and 4.
Unless the production function is known or there is a reference to an absolute profit maximizing production level, no change in management or farming practice can claim to distinguish between moving a production function and having moved to a different point along an existing production function. With that inherent ambiguity may rest a partial explanation for much of the confused information guiding dairy industry management and investment.
If there is no awareness of the difference between moving along a production function and moving the function itself, or if it is not possible to separate the effects of one from the other, confusion will reign.
A look at Figure 8 suggests that there are changes that would move a farm’s production function: Towards the right and higher milk solids production - improved pasture utilisation or higher dry matter production and; Others such as better herd management that move it up to higher operating surplus even though decreasing production.
Operating surplus is more useful in consideration of the management of an existing system than in appraising investment decisions on an existing or new production system. Operating surplus excludes finance and capital costs or accounting adjustments such as tax or depreciation and has similarities to earnings before interest and tax (EBIT) figures. Analysis can be undertaken to adjust operating surplus to give a standardized EBIT format.
Operating surplus is useful to provide a rapid comparison between similar average farm types or benchmarks and has the advantage of providing a simple comparative basis for different farm types – using the same resources – as stand-alone or integrated systems.
The milk solids payouts used as part of determining operating surplus intentionally reflect both the commodity milk price and the value add components of the payout from Fonterra. This basis represents normal farmer attitudes to payout. Any value-add component is though in effect a dividend and should be deducted from revenue for a true assessment of operating surplus for any given level of payout.
The use of a $4.24 milk solids payout for changes moving production along a production function (Figure 3 and Table 1) reflects the season past. The use of a $5.50 milk solids payout for moving production functions by changing management systems (Figure 8 and Table 2) reflects the season in prospect.
Pasture Profiles were derived working back from accounting data. These combined with lactation profiles are shown for the Waikato and Southland regions in Figures 1 and 2.
In considering Figures 1 and 2 it should be noted that maximizing farm operating surplus on pastoral dairy farms requires identifying the best match of herd size and calving date to dry matter production.
For the Waikato we have modeled production of 12,150 Kg of dry matter averaging 11 MJME (10.6-11.4) and with 77% utilisation. This level of dry matter production reflects average milk solids production in the Waikato, but is significantly less than the pasture dry matter production generally perceived.
Averaging data from Dexcel for the 21 out of 25 locations in the Waikato using the cage method of assessing dry matter and not using nitrogen provides a figure of 14.3 tons of dry matter per ha per annum.
Dexcel suggests that the cage method may reflect up to 30% higher dry matter production than the whole farm method due to the former reflecting the increased dry matter production from regular grazing. The Dexcel site had little recent information on dry matter production, and none for the same location reflecting the effect of nitrogen against a control.
The Impact of Nitrogen Fertilizers
There is a lack of precise information on dry matter response rates to the application of nitrogen fertilizer to pasture. Response rates are also likely to be localized and dependent on a number of factors including soil, climate and past nitrogen use.
Nitrogen responses will almost certainly conform to a law of diminishing returns making it difficult to be confident a) that the responses modeled are being achieved and b) the actual net utilisation achieved meets expectation.
Use of nitrogen is highly interchangeable with other means of tactical dry matter supply and its benefit is very sensitive to price, the intended purpose for the additional dry matter produced (direct grazing versus a boost for supplements) and the response rate achieved.
Nitrogen use was modeled at 2006 prices.
Nitrogen application increased operating surplus from higher milk production but with higher costs. With an optimisation model such outcomes are chosen as a function of the model selecting any increase in operating surplus. The small increase in operating surplus from the use of nitrogen could be viewed as not practically useful in a whole farm situation especially when considered against the resulting environmental impacts.
The nitrogen use modeled likely exceeds the level economically rational at the 2006 payout - and certainly given 2007-2008 costs.
A lower use of nitrogen would make little difference to the operating surplus modeled, simply decreasing cow numbers or shifting emphasis to other forms of tactical dry matter supply.
Regional Trends in dry matter and milk production
Dry matter quality and quantity appear to improve regionally moving from north to south. This alone may be responsible for milk production per cow also improving from north to south. Alternatively, the better match of pasture production with lactation in Southland over that for the Waikato evident from Figures 1 and 2 may be the major contributor.
A question remains around whether Canterbury fits the trend, and if the trend continues into latitudes higher than Southland.
12150 Kg dry matter versus expectations more in line with 14300 Kg dry matter
The curve in Figure 8 that stands out from the rest is the one reflecting dry matter production of 14,300 Kg per ha. This level of dry matter production may be in doubt, but could also be understated.
Adjacent farms with the “same” pastures may produce quite different dry matter quantity and quality figures. This is a function of management that has been illustrated on demonstration and monitor farms where flexible management policies apply (Ridler et al 1984).
Having a herd size greater than the optimal level reduces total operating surplus even if all animals are fed appropriately. Doing a good job of over producing is quite different from the situation of a farm simply being overstocked and not providing correct feeding to maximize animal performance. The latter is an inefficient production system that costs productivity and has a high negative impact on operating surplus. This latter situation and/or less than ideal grazing management practices due to excessive stocking at critical periods of the year may explain why the Waikato regional average farm appears to produce dry matter at a level lower than expected.
Establishing accurate regional levels of dry matter production would seem essential to understanding an apparent gap in interpreting dairy farm management performance.
Farm production functions for Waikato and Southland
NZ dairy production has moved to the right along their typical farm production functions shown in Figures 4 and 5. While not investigated in this assignment, parallel work suggests dairy farm production functions have hardly moved horizontally, but that they have moved vertically down in terms of operating surplus as costs have risen.
Figure 4 makes it clear that on the Waikato regional average farm any cow beyond a herd size of 288 will reduce the operating surplus - for any milk solids payout between $4.50 and $6.50 per Kg of milk solids with 2006 costs. Figure 6 suggests that the 2006 average herd size for the Waikato region of 305 cows would be optimal for a milk solids payout of $7.00 per Kg.
Figure 5 indicates how the Southland regional average farm’s herd size should respond to changing milk solids payout. Figure 7 suggests that the 2006 average herd size of 445 cows would be optimal for a milk solids payout of $6.25 per Kg – at payouts less than this operating surplus would improve from reducing herd size. Figure 5 shows little marginal return in increasing herd size beyond 424 cows even at a milk solids payout of $6.50.
There is clearly a good return from dairy farms reviewing their levels of production against information from their farm’s production function. Simply relying on additional cows to provide improved operating profit is no longer a viable option. Costs of additional dry matter and the (apparent) current pasture production/utilisation “ceiling” provide very “flat” figures for optimal herd size about normal payout levels.
Regional differences between Waikato and Southland
Farm production functions for Waikato and Southland show similar characteristics but those for Southland are better positioned. Higher per cow production in Southland provides a lower cost structure resulting in a higher operating surplus and requiring a lower payout to breakeven. This can be seen comparing changes in the Waikato and Southland operating surplus in Figures 6 and 7 as the milk solids payout decreases.
It appears pasture and cow lactation profiles are better matched in Southland. This combines with higher quality pasture to provide higher efficiency of conversion of pasture to milk. This is improved further by the efficiencies that flow from higher production per cow (dilution of maintenance overheads).
Setting the correct Production Level
Setting the correct production level was the key management decision engaging dairy farmers from the late 1950’s through till the 1980’s - for good reason.
Table 1 illustrates clear differences between current regional average and resource optimised regional average farm performance across all regions, but with a trend in operating surplus gains from south to north.
The total effect of all regions minus Canterbury operating at an optimal production level would have been a gain in operating surplus of $252 million in 2006, but coupled with an 8.4% reduction in milk solids production. This would have been an average increase in operating surplus of 24% or 22.7 cents per Kg of milk solids. The gain per Kg of milk solids would have been highest in Northland (35 cents), followed by Waikato (24 cents), the Lower North Island (20 cents) and Southland (13 cents).
These figures are for 2006 costs and a payout of $4.24. The 2008 season with a much higher milk solids payout will likely see production increases attempted to a greater extent. The demand for sources of additional dry matter will increase costs. The net effect on operating surplus of overproduction driven by a higher payout at this point remains unclear.
In Figure 8 all curves include the production level that maximizes operating surplus and which shows as the peak in the curve. All these management changes improve operating surplus over typical management - some considerably. Particularly interesting is that better herd management both decreases production and increases operating surplus.
None of these management changes were explored in depth as part of this project. This preliminary analysis should encourage further work in this area.
An area of management not investigated involves seasonal aspects of time. Dry matter has a marked time value, and adjusting the timing of calving, the making of silage or the application of nitrogen can produce marked changes in levels of production and operating surplus. This aspect of management equally warrants further investigation.
Summary of Discussion
Regional models will be less precise than the individual farms making up the region. Answers to questions being raised by differences between regions will likely be explained by analysis of individual farms.
The use of regional averages tells us little about the distribution of farm performance - that remains unclear. The performance of individual farms would tell us much more.
Many NZ dairy farming systems are now producing close to or beyond the point where MC=MR. That point on pastoral farms is defined by a major break – a sharp increase in costs as the farm switches to higher intensity production and the use of supplements. It is incorrect to spread these costs over all the past production, or to assume that these feeds have gone solely to increased production when increasing pasture intake at critical periods may well have had the same effect.
There is a discontinuity between high performance dairy systems and lower intensity pastoral ones that does not appear to favour any form of hybrids. Risk and return around the point where MC=MR remains a fascinating area of analysis.
High-energy supplements are often fed in an attempt to increase per cow production on high performance dairy systems. Although the milk production possible from pasture has been shown to be between 400 and 420 Kg of milk solids per cow (500 Kg live weight), this is rarely achieved in practice.
The first change should be management to achieve the potential from grass alone. This will often require a better fit of pasture to animal requirements and may require a change in calving date or a reduction in stocking rate rather than filling in feed shortages with expensive supplements.
Pastoral farming systems are cost effective but complex to understand and manage at the margins. In the past NZ dairy farms were not operated in close proximity to these margins and were easier to understand and manage.
Knowing the nature of the basic production function for pastoral dairy farms is fundamental to understanding the NZ dairy industry. Knowledge of their farm's individual production function will be advantageous for farmers operating close to where marginal cost exceeds marginal revenue in setting production levels for maximum profit.
Production functions are unique to each farm and also for each time period. The use of industry or regional average production functions instead of individual ones may be of little benefit beyond educational.
Traditional approaches to analyzing farm performance are ex post analysis working from outcomes back to expected capacity rather than ex ante analysis forward from capacity to maximum profit.
Information such as production functions has never before been available to farmers. Instead farmers have had to make production decisions based on a large amount of conflicting information and often confusing market signals.
Dairy farmers have choices between less profit from more production moving to the right along current production functions (operating surplus curves), more profit and more production from moving production functions up and to the right, or more profit from less production by moving production functions up and to the left.
In practical terms, maximizing farm operating surplus on pastoral dairy farms, in order of priority, comes down to:
The potential exists for dairying to increase profitability while simultaneously reducing societal and emissions costs, but issues remain over farm management’s willingness to respond to information. How will dairy farmers respond if provided with information for their farm equivalent to that in Appendices 1, 2, 3 and 4?
There are significant gains for NZ dairy farming from Resource Allocation Optimization - $250 million in 2006 – 23 cents per Kg milk solids or 24% gain in operating surplus. Associated with this uptake may be cultural and attitudinal shifts more significant than the financial gains.
Bio-economic models are effective tools capable of providing a wide range of information from simulations. More extensive use would be expected of these tools in farm management as modeling continues to clarify the information required for improved farm decision-making, and the presentation of that information gains a wider audience.
Working with regional average farms the GSL bio-economic model appears to produce sound projections. The only issues that may be of concern with results to date are those of accurate calibration of MAF average farms against commercial farm systems and the accuracy of some fundamental input data such as nitrogen response rates. Further modeling with commercial farm systems should be undertaken to confirm neither is an issue.
The direction of further agricultural production economic modeling will need to evolve in wider discussions including establishing the various contexts of different parties and any international comparisons required regards competitive advantage. Any modeling is likely to additionally require aspects of monitoring and the communication of findings to a number of different audiences.
This and concurrent work has clarified the information possible from simulation modeling, and the form and presentation of such information required to allow more efficient investment or resource allocation decisions. Continuing this work with the following is suggested:
Dexcel - Average pasture growth data for New Zealand dairy farms:
Colin Riden, 2007. The Marginal Cow I – Allocative Efficiency:
Colin Riden, 2007. The Marginal Cow II – Productive Efficiencies:
Ridler BJ, Hurley E, Stachurski LJ; 1984: The management policy on Massey University’s Large Herd. Proceedings of Ruakura Farmers Conference.
GSL and the GSL Model refer to GSLdiagnostic software from Grazing Systems Ltd,
17 Chilton Grove, Palmerston North.
For background on the use of LP in Farm Management System Models refer to:
Ridler, BJ; Rendel, JM: Baker A; 2001 Driving Innovation: Application of Linear Programming to improving Farm Systems. Proceedings of the New Zealand Grasslands Association pp295-298
The link to: Dexcel - Average pasture growth data for New Zealand dairy farms is no longer available, but an alternative is:
Four graphs present data from the same model but in different combinations hoping one or more will communicate the critical importance in pastoral dairy farming of setting the correct production level (MC=MR for any potential milk solids payout).
Appendix 1: Marginal Cost per Kg of milk solids plus Supplemental Dry Matter used.
Marginal Cost per Kg of milk solids plus Supplemental Dry Matter used. The steep increase in marginal cost with the introduction of purchased supplements (maize silage) is the critical point to note.
Comments on this article can be made by any athenticated user. A discussion forum is also available: