All models are wrong, some are useful!

Home / Articles & Publications / Opinion Pieces / All models are wrong, some are useful!

This astute aphorism is attributed to George Box, a British statistician of the twentieth century. It is something that any of us who perform modelling or longer-term budgeting should keep in mind.

A phone call from a beef producer to discuss recent extension messages on herd fertility reminded me of the quote. The producer was attending an event where one of the key messages being delivered was that it was not economical to pursue improved fertility via genetics in northern beef herds. This surprised her as it was contrary to her own experience, most other extension messages and the findings from our analyses over time.

The messages were from a Qld Dept. of Agriculture and Fisheries report on preparing for, responding to, and recovering from drought. This specific report was for northern Gulf beef production systems, one of a series of similar reports produced across regions and over time.

The reports are all underpinned by analyses conducted using Breedcow Dynama herd budgeting software, a complex but useful herd model. The analyses apply sound economic theory to herd investment decisions by accounting for diminishing marginal returns, comparing the marginal cost of an intervention with the marginal benefit, and discounting future cashflows. Broadly, this is an analytical approach we support and these concepts and calculations are taught in the BusinessEDGE workshop.

Breedcow Dynama models herd performance based on scenarios and assumptions entered by the user. With a skilled operator who understands the model, its intricacies, herd dynamics and who can sense check the results, it is a powerful tool. However such users are few and far between.  

Like all models, Breedcow Dynama is nothing more, or less, than a set of calculations that attempt to model or reflect reality. However, models can never fully reflect reality. They will only be as accurate as the calculations they are made up of and the assumptions that are applied to them, as George Box alluded to.

Within the Breedcow model, particular attention needs to be paid to when females are assumed to wean their first calf, whether weaners are produced from culled cows, the sales strategy for surplus females, input costs and when during the course of the year mortalities occur. Consideration also needs to be given to the fact that the relative feed intake of the various classes is calculated in a simplistic manner so comparisons involving different turnoff weights and herd profiles are not necessarily apples and apples in terms of total grass consumed.

Some of the key messages regarding fertility from this northern gulf report are;

  • the returns from investing in genetics to improve fertility are not inviting and it is a risky investment,
  • using home bred bulls provides better returns than investing in better genetics, and.
  • investments in improving herd performance (other than P supplementation) are unlikely to substantially improve profit

I disagree with these messages; I do not think the modelling results or their interpretation reflect reality.

In some ways they seem to contradict general findings from a large scale study done in the same region twenty years earlier using the same software[1]. This study found a general improvement in branding rates [and decrease in death rates] was the key for many properties. A more recent study of actual business data in the region[2] found a correlation between the herd productivity drivers and business profitability.

We have analysed many hundreds of business years of data from actual northern beef businesses and have spent a lot of time interpreting these data to understand which factors separate the highly profitable beef producers from the rest.

Consistently at the top of the list is that the better performing beef businesses have higher herd productivity (i.e., they are more efficient at turning grass into beef) and the number one productivity driver for breeding businesses is reproductive rate, or herd fertility. The others productivity drivers are sale weight and mortality rate. These three elements explain three quarters of the difference in productivity between herds. Productivity (kilograms of beef produced per animal unit) explains nearly all the difference in income between herds over time as there is minimal relative difference in price received. Price received is also not a factor separating the best from the rest in the long term.

There are two interrelated components to herd fertility; management and genetics. Genetics, while important, are usually secondary to management of the beef businesses herd and pastures.

A key component of management is how females are selected and managed. Selecting the most fertile of the younger females coming through accelerates genetic gain. Mating all heifers allows the most fertile ones to identify themselves. Maximising the sale weight of cull females will further compound the gains and improved reproductive performance presents options for value adding to female sales such as selling cull animals as PTIC or a cow & calf unit.

There has been some good work done by the NT primary industries department over time on quantifying the gains that can be made through selection and management to herd fertility and genetics[3]. In this work females were selected for fertility and bulls selected from within the herd for breeding soundness and using Breedplan data for fertility. Far greater improvements in reproductive rates were achieved in this trial than modelled in the above-mentioned report (35% higher pregnancy rate vs. 5.5% higher weaning rate), over a similar time frame (~12yrs).

Improving sourced genetics should not be looked at, or modelled, in isolation from management. In my experience, top performing managers have an unrelenting focus on the productivity drivers and use genetics as a tool to leverage their herd and business management to achieve high performance. The skill and tenacity of a good manager is a difficult variable to model.

Improving fertility will improve business performance for nearly all northern beef businesses, but it may not be the priority lever for some to pull. It is important for producers to understand where fertility ranks as a constraint to their performance, I’d say it will be in the top 3 for most. It may be that turnoff weight, mortality rate, operating scale, labour efficiency, land condition and/or debt levels are more of a constraint on their business performance. Each business is unique; one size does not fit all.

Breeding your own bulls is a strategy that can work, as evidenced by the NT work mentioned above, but it must be done well. A practical strategy I’ve seen work is using your own proven females (those that have produced a weaner each year from at least their first two joining’s) to ‘multiply’ externally sourced genetically superior bulls.

As with all improvements, it is uneconomic to spend more on the improvement than it actually returns but, in the case of buying bulls, it is just as easy to spend too much on a bull that is unlikely to improve genetics as it is to overspend on one that will. Genetic improvement is persistent and compounding, but you must acquire the right bulls at the right price to make it profitable. A lot of the management interventions also come at either no cost or are cash positive, e.g. through selling unproductive females.

There is a plethora of quantitative business data, research findings, and even modelling exercises which show that significant gains in fertility can be made (see list below), and that improved herd fertility and productivity is associated with better business performance, even in challenging northern environments. The question is not whether it is economical to improve herd productivity. It is how is herd productivity increased to improve the economic performance of northern beef businesses.

There is a definite place for modelling; it and actual business data (empirical evidence) can complement each other. Neither are perfect in their own right. However, if a modelling exercise has an outcome that is vastly contrary to the empirical evidence, then the modelling, its underlying calculations and applied assumptions should be checked and ground-truthed. The modelling results should not be treated as fact. Nor should they form the basis of recommendations to producers.

Suggested Reading

ASH, A., HUNT, L., MCDONALD, C., SCANLAN, J., BELL, L., COWLEY, R., WATSON, I., MCIVOR, J. & MACLEOD, N. 2015. Boosting the productivity and profitability of northern Australian beef enterprises: Exploring innovation options using simulation modelling and systems analysis. Agricultural Systems, 139, 50-65.

HARBURG, S., MCLEAN, I. A. & HAYES, B. J. 2020. Economic consequences of selection for cow fertility in northern Australian Beef Herds. Queensland Alliance for Agriculture and Food Innovation.

HOLMES, P., MCLEAN, I. & BANKS, R. 2017. The Australian Beef Report, Bush AgriBusiness.

KERNOT, J., ENGLISH, B. H., HILL, F., LAING, A. & SMITH, P. 2002. Smart Manager. Project number: NAP3.107. Meat & Livestock Australia Limited.

MCCOSKER, T. H., MCLEAN, D. K. & HOLMES, P. R. 2009. Northern beef situation analysis 2009. Meat & Livestock Australia.

MCLEAN, I., HOLMES, P., COUNSELL, D., BUSH AGRIBUSINESS & HOLMES & CO 2014. The Northern Beef Report. North Sydney: Meat and Livestock Australia.

MCLEAN, I. A., HOLMES, P. R., WELLINGTON, M. J., WALSH, D. L., PATON, C. J. & FREEBAIRN, R. D. 2020. The Australian Beef Report: 2020 Vision.

ROLFE, J. W., LARARD, A. E., ENGLISH, B. H., HEGARTY, E. S., MCGRATH, T. B., GOBIUS, N. R., DE FAVERI, J., SRHOJ, J. R., DIGBY, M. J. & MUSGROVE, R. J. 2016. Rangeland profitability in the northern Gulf region of Queensland: understanding beef business complexity and the subsequent impact on land resource management and environmental outcomes. The Rangeland Journal, 38, 261-272.

SCHATZ, T. J., JAYAWARDHANA, G. A., GOLDING, R. & HEARNDEN, M. N. 2010. Selection for fertility traits in Brahmans increases heifer pregnancy rates from yearling mating. Animal Production Science, 50, 345-348.

[1] Kernot et al., 2002

[2] Rolfe et al., 2016

[3] Schatz et al., 2010

Leave a Reply