Free access

Benefit-cost analysis of near-infrared spectroscopy technology adoption by Alberta hog producers

Publication: Canadian Journal of Animal Science
25 March 2020


Feed cost is a significant component of livestock production costs, accounting for over half of total operating costs for hog producers. This provides an incentive to minimize feed costs while meeting dietary requirements. However, producers may not know the nutritional content of their feed grains with certainty. Near-infrared reflectance spectroscopy (NIRS) imaging technology can quickly and accurately estimate the nutritional content of different types of feed grain. Although the technology has been available for almost five decades, producer adoption has been slow due to issues with cost and usability. The objectives of this study are to estimate feed cost savings resulting from the adoption of NIRS on a representative Alberta hog farm and to conduct a benefit-cost analysis to investigate the potential viability of NIRS adoption. A joint mathematical programming-simulation approach is used to estimate the cost savings generated by adoption of NIRS technology. Results suggest mean annual savings of up to $4 per hog and benefit-cost results suggest that adopting NIRS technology may be viable, particularly for larger Alberta hog operations. However, initial investment requirements, uncertainty in the magnitude of benefits, and access to the technology from feed mills will likely continue to limit adoption.


Les coûts d’alimentation sont une composante significative des coûts de production du bétail, et représentant plus de la moitié des coûts totaux de fonctionnement pour les producteurs de porcs. Ceci offre une motivation pour minimiser les coûts d’alimentation tout en répondant aux exigences alimentaires. Par contre, les producteurs pourraient ne pas connaître le profil nutritionnel des grains alimentaires avec certitude. La technologie d’imagerie de spectroscopie de réflectance dans le proche infrarouge (NIRS — « near infrared reflectance spectroscopy ») peut rapidement et précisément estimer la teneur en éléments nutritifs de différents grains alimentaires. Bien que la technologie soit offerte depuis presque cinq décennies, l’adoption par les producteurs a été lente pour des questions de coûts et de convivialité. Les objectifs de cette étude étaient d’estimer les économies de coûts d’alimentation résultant de l’adoption de la NIRS sur une ferme représentative de production de porcs en Alberta, ainsi que d’effectuer une analyse coût-bénéfice pour évaluer la viabilité potentielle de l’adoption de la NIRS. Une approche conjointe programmation mathématique et simulation a été utilisée afin d’estimer les économies générées par l’adoption de la technologie NIRS. Les résultats suggèrent des économies moyennes annuelles jusqu’à 4 $ par porc, et les résultats de l’analyse coût/bénéfice suggèrent que l’adoption de la technologie NIRS pourrait être viable, en particulier pour les plus gros producteurs de porcs en Alberta. Par contre, les investissements initiaux requis, l’incertitude par rapport à l’ampleur des bénéfices, et l’accès à la technologie pour les usines d’alimentation continuera vraisemblablement à limiter l’adoption de la technologie. [Traduit par la Rédaction]


The cost of feed is a significant component in the livestock industry’s cost of production. For example, depending on the sourcing of ration ingredients (purchased versus home mixed), total feed costs for a representative 500 sow Manitoba farrow-to-finish operation may represent up to 79% of total operating costs (MAFRD 2018). Feed prices and costs continue to increase for hog producers; for the same representative 500 sow Manitoba farrow-to-finish operation, total feed costs increased by 28% from 2015 to 2018 (MAFRD 2016, 2018). From 2014 to 2019, prices for feed barley and feed wheat (two important hog ration ingredients) in Alberta increased by 34% and 23%, respectively (AAF 2019). As a consequence, hog producers have a strong incentive to minimize feed costs as they account for well over half of total operating costs.
In general, smaller hog producers source their feed from commercial feed mills or they may grow their own feed. Commercial feed mills purchase grains from farmers, use the grains to manufacture feeds (with or without the addition of feed additives), and in turn sell the feed to hog producers. Larger hog farms may operate their own feed mills and purchase grain directly from crop farmers. Hog feed is composed mainly of grains and hogs consume between 35% and 45% of Canadian feed grains (CPI 2013). The hog’s diet is typically composed of corn (54%), barley (29%), and wheat (12%), whereas other types of feedstock such as canola meal and soybean meal are used to supplement feed grains (CPI 2013). Although most Canadian corn is produced in central Canada, barley and wheat are predominantly produced in the Canadian Prairie region. Therefore, hog diets in the Prairie region tend to be more wheat and barley intensive than those in central Canada; for example, the rations provided by MAFRD (2018) for the representative Manitoba farrow-to-finish operation, and the control finishing ration used in the study by Zhang et al. (2016), make extensive use of wheat and (or) barley.
To be profitable, producers need to keep their feed costs as low as possible subject to meeting their animals’ dietary requirements. What makes this constrained optimization problem challenging from a practical perspective is that producers often do not have complete information regarding the nutritional content of their feed. This can lead to an inefficient outcome arising from either overfeeding or underfeeding animals. Providing insufficient nutrients can extend the time to attain market weight or result in hogs being marketed at a sub-optimal weight. Conversely, providing excess nutrients results in feed wastage and higher costs. Having accurate and complete knowledge of nutrient content for ration ingredients would lead to cost savings and higher profits. This is consistent with the use of precision feeding of hogs (Pomar and Remus 2019), where research shows that greater precision in meeting nutrient requirements for growing pigs results in cost savings as well as environmental benefits (e.g., reduced phosphorus excretion and greenhouse gas emissions).
Near-infrared reflectance spectroscopy (NIRS) imaging technology can quickly and accurately estimate the nutritional content of different types of feed grains. The NIRS technology has been available for almost five decades, but livestock producers have been slow to adopt the technology due to cost and usability issues. However, recent technological advances have resulted in a more compact and affordable machine. Adoption of NIRS at the farm level can increase the likelihood of providing the most cost-effective ration thereby resulting in higher profitability, but this depends crucially on the relative benefits and costs of adoption.
The main motivation for adopting this technology is the ability to calibrate rations more precisely, which in turn can lead to a more cost-efficient feeding program and ultimately higher profit. The main barrier to adoption is the high entry cost; that is, the initial purchase price. Only a handful of the larger hog producers in Alberta have adopted NIRS machines for example, whereas the vast majority of Alberta producers have not adopted the technology due to the prohibitive cost of the machine and the maintenance of calibrations.1 Other potential barriers to adoption include the learning curve associated with adopting a new technology and the uncertainty of potential benefits.
Although there has been extensive research on the technical viability of NIRS technology (Norris 1996; Cozzolino and Moron 2004; Pujol et al. 2007; Modroño et al. 2017), there is a gap in the literature with respect to the economic benefits of adoption. The main objective of this paper is to examine the economic viability of adopting NIRS. To that end, this paper has two goals: (1) estimate the feed cost savings resulting from the adoption of NIRS on representative Alberta hog farms and (2) conduct a benefit-cost analysis to investigate the viability of NIRS adoption.

Materials and Methods

The empirical results reported in this paper are generated using a combination of simulation analysis and constrained optimization techniques. The analysis is done for representative feeder-to-finish hog producers, who are assumed to minimize feed costs subject to nutrient requirement constraints that support pre-determined rates of weight gain for the various stages of growth. Nutrient requirements for each stage are known with certainty, but nutrient content is assumed to be uncertain, with nutrient content in ration ingredients being independently and normally distributed.
A simplifying assumption made in modelling uncertain nutrient content here is that only crude protein content in select ingredients (i.e., peas, barley, and wheat) is uncertain. The main reason for focusing on crude protein is that, while NIRS measures protein and starch content, data on starch content of key feed ingredients are much more limited in availability.2 Other potential ration ingredients also contribute protein (e.g., canola meal or soybean meal), but they are assumed to have undergone significant processing, and so the protein content is more certain. It is further assumed that the distribution of crude protein content can be adequately characterized by the mean and variance.
The baseline scenario represents the case in which NIRS technology is not adopted by the hog producer. Under this scenario, hog producers are aware that crude protein content in peas, barley, and wheat is uncertain. However, other than knowing the nature of the distribution for crude protein, they have no other ex ante information about nutrient content. Further, hog producers are assumed to be averse to under supplying nutrients to their hogs. To represent this scenario, a constrained cost minimization model is constructed to formulate rations, allowing for a “safety” buffer that ensures nutrients levels are met with a predetermined level of confidence. If nutrient content is assumed to be normally distributed, then a linear programming margin of safety (LPMS) model can be formulated following D’Alfonso et al. (1992). The general LPMS specification is
Subject to
In the linear programming formulation represented by eq. 1, the objective is to minimize total ration cost, where X represents the vector of ration ingredient quantities, and W is the vector of ingredient unit prices. The first constraint set represents requirements for all nutrients except for crude protein, where A is a matrix of nutrient content values with individual elements aij being the amount of the ith nutrient provided per unit of jth ingredient, and b is a vector of nutrient requirements per animal. The second constraint set incorporates any relevant restrictions on ingredient use in the ration, where C is a matrix of coefficients for feed ingredient restrictions, and d is a vector of ingredient limits in the ration. The final constraint represents the LPMS component, ensuring that the crude protein requirement bp is met with probability α. In this constraint, μpj and σpj are the mean and standard deviation, respectively, of crude protein content in the jth feedstuff, and Zp is the standard normal score associated with α.
The alternative scenario modelled in this analysis is the case where the producer adopts NIRS technology. With the adoption of NIRS, the uncertainty associated with nutrient content (i.e., crude protein) is assumed to be eliminated; that is, all nutrient content for all ration ingredients is known with certainty. In this case, the ration formulation model is simplified to a standard least cost ration (LCR) model, as follows:
Subject to
where the model is defined as before for the LPMS model in eq. 1, with the exception that now the crude protein requirement constraint is defined as for the other nutrient requirements; that is, apj is the (certain) crude protein content for the jth ingredient, and bp is the crude protein requirement per animal.
To empirically capture the impact of the variability in crude protein nutrient content, repeated sampling is done from pre-determined crude protein distributions for the relevant ration ingredients. The distribution parameters used in the sampling are the same as those incorporated in the linearized chance constraint specified in the LPMS model (i.e., eq. 1).
For the NIRS adoption scenario, the LCR model in eq. 2 is resolved for each draw so that a distribution of LCRs and associated minimum ration costs is generated. For the base (non-NIRS) scenario, the solution from the LPMS model (eq. 1) is reevaluated to determine whether the required level of protein would actually have been provided by the optimal LPMS ration. For those sample draws where the crude protein provided by the uncertainty model optimal ration falls short of the hog’s requirement (i.e., the crude protein constraint is violated), the effect on carcass weight due to the nutrient deficiency is estimated. This is then used to estimate the loss of lean weight premiums (premium penalties) earned in selling the hog carcass. This sample and solve procedure is repeated 1000 times.

Hog enterprise characteristics

For both baseline and NIRS scenarios, the ration formulation models are solved for seven hog growth stages. The stages represent seven levels of hog development with the first stage being newly weaned pigs and the last stage being finishing pigs to be marketed. Hogs in the different growth stages have different weight ranges and growth rates. Therefore, nutrient requirements will vary between stages.3 For the purposes of the analysis in this paper, a feeder-to-finish hog enterprise is assumed meaning that each hog on the farm goes through each of the seven growth stages. It is further assumed that the hogs are uniformly distributed across the different growth stages. This allows for straightforward aggregation of total feed cost across stages to represent a daily feed cost for a representative hog. The daily feed cost is then annualized and aggregated to the farm level for two sizes of commercial hog farms: 3453 head and 13 327 head, referred to as small and large, respectively.4

Benefit-cost analysis

To assess the financial viability of adopting NIRS technology, farm level cost benefit analysis is done using net present value (NPV) calculations. The daily net benefit of adopting NIRS is computed by calculating the difference in the daily feed cost per hog between the baseline (no NIRS) and alternative (NIRS adoption) scenarios, incorporating applicable premium penalties as noted earlier. This is done for each nutrient content draw thereby yielding a distribution of net benefits associated with the adoption of NIRS technology. The median net benefits of adopting NIRS are then converted to an annual value and aggregated to the farm level for the small and large hog operations.
To examine the viability of investing in (adopting) NIRS technology, the cost of procurement must also be considered. There are two main types of NIRS machines or analyzers: (1) an inline or process analyzer and (2) a sample or desktop analyzer. The process analyzer is ideal for measuring nutrient content “on-the-fly” and captures the nutrient content of the entire grain delivery. The desktop analyzer accurately measures the nutrient content of representative samples of the entire grain delivery.5
The annual net benefits are discounted and then summed to arrive at a present value equivalent. The initial procurement cost is then subtracted to obtain the NPV. The formal calculation is as follows:
where INV is the initial capital investment in NIRS technology, NBt is the annual farm-level net benefit from adoption in year t, and i is the discount rate. If the NPV is non-negative, the investment in NIRS technology is financially viable; that is, it earns a rate of return sufficient to justify the investment.
For the NPV calculation, a discount rate and investment time horizon are required. The discount rate should reflect the opportunity cost of capital. In this study, discount rates of 5%, 8%, and 10% are used to reflect the range of public (5%) and private (10%) opportunity costs of capital. The time horizon should reflect the useful lifespan of the machine. Based on expert opinion,6 the estimated life of a NIRS machine is approximately 20 yr, and so that is used as the time horizon for the NPV calculations.
Sensitivity analyses are conducted to estimate the impact of changing key parameters used in the NPV calculations. In particular, sensitivity analysis is done on key ration ingredient costs, the discount rate used to calculate NPV, and the impact of relaxing the probability of meeting the nutrient requirements on net benefits.


Potential ingredients to be incorporated into the LPMS and LCR models were selected based on expert opinion. The primary source of ration ingredient prices was MAFRD (2018). Other published sources were used in cases where prices for relevant ingredients were not available from that source. A list of feed ingredients used in the ration modelling, their associated prices, and the sources of those prices is provided in Appendix Table A1.
For the three ingredients assumed to have uncertain crude protein content (i.e., wheat, barley, and peas), the actual protein content is independently and normally distributed. The parameters of these distributions are used to parameterize the linearized protein chance constraint in the LPMS model, and also for the draws of actual protein levels used in optimizing rations in the NIRS scenario and determining whether crude protein requirements are violated (or not) for the baseline (non-NIRS) scenario. In the baseline scenario, the LPMS model is solved for the optimal ration given a predetermined probability (95%) of meeting the nutrient requirement. As noted earlier, sensitivity analysis is done for the value of this probability. The means and standard deviations provided in Table A3 are used in the first constraint of eq. 1.
Nutrient requirements and the maximum allowed levels of ration ingredients for each production stage are provided in the Appendix in Tables A2 and A4, respectively. In the case of limits on amino acids, crude protein is assumed to be linearly related to amino acid content in specific ration ingredients (NRC 1998). The linear relationships are factored into the LCR and LPMS models according to the specifications from the nutrient requirement of swine nutrition guide (NRC 2012). The linear relationships and data used to characterize the variability of protein content are provided in Fig. A1 and Tables A3 as well as A5.
Investment and maintenance costs associated with procuring and using NIRS technology are obtained from Foss North America (a NIRS analyzer manufacturer). For the process analyzer, the investment and annual maintenance costs are $143 649 and $10 567, respectively, whereas for the sample analyzer, the costs are $32 648 and $6650, respectively. The maintenance costs incorporate the cost of calibrating the machines.
Data used to model the growth impact of protein deficiency are obtained from Campbell et al. (1985). Regression analysis is conducted to estimate the marginal impact of protein intake on the protein weight deposition in the carcass (see Table A6 in the Appendix). The protein weight loss is then valued using the estimated change in representative price premiums paid by the Western Hog Exchange (e.g., Table A7 in the Appendix) in relation to the change in carcass weight categories.


The empirical results are presented first in terms of the cost impact of NIRS technology, followed by a presentation of the NPV results. Lastly, results from the sensitivity analysis are presented.

Estimated ration cost impact of NIRS technology adoption

The baseline (non-NIRS) scenario results presented in Table 1 indicate that daily ration costs per head increase as the hog progresses through the seven growth stages; from $0.11 for a starter (stage 1) to $0.83 for a finisher (stage 7). However, despite the increase in costs per hog over the growth stages, the per kilogram cost of the rations decreases and then is relatively stable. The total cost to feed a representative hog through all seven stages is $95.51 (Table 1).
Table 1.
Table 1. Summary output of the baseline scenario least cost ration models (no near-infrared reflectance spectroscopy).
Under the NIRS adoption scenario, the sample and optimize protocol yield a distribution of costs for each stage. A summary of the resulting cost distributions for the seven growth stages is presented in Table 2. Similar to the LPMS results for the baseline scenario, the mean daily ration costs generated from the LCR models used for NIRS adoption also increase significantly through the seven growth stages. The mean cost for the representative hog is $93.04 (Table 2).
Table 2.
Table 2. Summary simulation results for linear program margin of safety model [near-infrared reflectance spectroscopy (NIRS) adoption].
Figure 1 illustrates the distribution of annual cost savings per hog associated with adoption of NIRS. If the effects associated with violating the protein requirement constraint are not considered, the mean cost saving is approximately $2.00 per hog. As shown in Fig. 1a, approximately 23% of draws (231 of 1000) result in cost savings greater than $4 per hog, whereas approximately 20% (208 of 1000) result in zero or negative cost savings from adoption of NIRS.
Fig. 1.
Fig. 1. (a) Distribution of cost savings from adoption of near-infrared reflectance spectroscopy (NIRS) (no premium penalty included) and (b) distribution of cost savings from adoption of NIRS (premium penalties included). [Colour online.]
As shown in Fig. 1b, if the impact of violating nutrient content is considered then the distribution changes significantly. The mean annual ration cost savings per hog increases to $3.95, and the chance of incurring higher costs due to increased information about nutrient content falls to zero (0 out of 1000 draws), whereas the likelihood of savings greater than $4 per hog increases to almost 48% (477 of 1000 draws). If a margin of revenue over variable costs of between $16 and $36 per hog were assumed (MAFRD 2018; depending on whether home-grown or purchased feed is used), these results suggest that cost savings from NIRS adoption could expand margins by between 10% and 25%.

Results for NPV

A summary of the distribution of NPV results for NIRS adoption, assuming an 8% discount rate, is provided in Table 3. As noted earlier, positive NPV values may be interpreted to mean that adoption of NIRS would result in increased wealth or welfare for the hog producer.
Table 3.
Table 3. Summary of net present value (NPV) analysis for near-infrared reflectance spectroscopy adoption (8% discount rate).
For the case of the small farm (3453 hogs), the results presented in Table 3 suggest that investment in NIRS technology may be economically feasible, but this is not certain. The mean NPV for the process analyser is negative (−$113 491), as is the 75th percentile value (−$64 630). In the case of the sample analyzer, the mean NPV is positive ($35 975). However, the 25th percentile value is negative (−$24 522), suggesting that for a significant number of simulated adoption scenarios, even the sample analyzer was not economically viable for the smaller hog operation.
Conversely, for the large hog farm (13 327 hogs) adoption of NIRS technology through investing in an analyzer does appear to be economically viable. Mean NPVs for both types of analyzers are positive; $418 893 for the sample analyzer and $269 427 for the process analyzer. In addition, although the minimum NPV value for both adoption scenarios is negative, the 25th percentile values are both positive, suggesting a high likelihood of positive economic benefits from adoption.
It should be noted that the analysis for the sample analyzer assumes that all nutrient content uncertainty is eliminated. In practice, this would not be the case given that only ingredient samples are analyzed, although uncertainty would be reduced relative to the no-NIRS scenario. Thus, these results should be interpreted as representing the upper bound on net benefits from adoption of this form of the technology. However, given the magnitudes of the mean NPV for the large-sized operation, it seems likely that for many larger producers the sample analyzer would be economically viable.
The pattern in results presented here is largely attributable to differences in farm size. Although there are significant cost savings per hog associated with the use of NIRS, the enterprise size for the smaller representative hog farm does not allow the producer to take sufficient advantage in terms of total cost savings to justify investment in the technology.

Sensitivity analysis

There are a number of key assumptions made in undertaking the analysis reported in this paper, some of which may impact on the results. The discussion here focuses on three of those assumptions: (a) unit prices for key ration ingredients, (b) the choice of discount rate in the NPV analysis, and (c) the probability of non-violation of the crude protein requirement constraint specified for the LPMS model. The impact of the CAD:USD exchange rate is also considered in this section.
The value of adopting NIRS technology may be dependent on the cost of potential ration ingredients. To examine this possibility, mathematical programming-simulation analysis is redone with unit prices for key ingredients being increased and then decreased; specifically, the prices for wheat, barley, and peas are jointly changed by 10% in both directions. These three ingredients were chosen due to their importance as the ingredients assumed to have uncertain crude protein content.
As shown in Table 4, the changes in ingredient prices result in a consistent pattern of change for the NPV results. With higher ingredient prices, the economic viability of investing in NIRS technology is weakened, whereas the opposite is true for lower ingredient prices. These results are not surprising. If ingredients with uncertain protein content become relatively more expensive, a smaller quantity of these ingredients will be used in the LCR (regardless of whether or not NIRS is adopted). This reduces the magnitude of benefits from adopting NIRS, either through direct cost savings or reduced premium penalties due to violations of the crude protein constraint. Conversely, if those ingredients become relatively less expensive, greater amounts will be used in rations leading to greater value in having more certain information about nutrient content.
Table 4.
Table 4. Sensitivity analysis of near-infrared reflectance spectroscopy adoption net present value (NPV) results (varying unit cost of wheat, barley, and peas).

Note: The 10% higher and lower scenarios involve jointly increasing or decreasing, respectively, the prices of wheat, barley, and peas.

As discussed earlier, the primary analysis done for this study uses a discount rate of 8%. This is chosen as a “midpoint” value based on a review of literature and consideration of social versus private opportunity costs. Table 5 provides a summary of the NPV results (mean values), varying the discount rate for the two hog operations and the two alternative types of analyzers. Besides the baseline rate of 8%, rates of 5% and 10% are also used to calculate NPVs. The 5% rate would be typical of a social discount rate, whereas 10% would be towards the upper end of private opportunity costs. Not surprisingly, increasing (decreasing) the discount rate results in mean NPVs that are lower (higher). However, the qualitative patterns presented and discussed earlier are unchanged; that is, the sign for mean NPV associated with adoption of NIRS technology remains unchanged.
Table 5.
Table 5. Sensitivity analysis of near-infrared reflectance spectroscopy adoption net present value (NPV) results (varying discount rate).
Sensitivity analysis (Fig. 2) is also conducted on the specified probability of non-violation of the crude protein nutrient requirement constraint. The value of this probability can be interpreted as being a simplified representation of the degree of risk aversion displayed by the producer with respect to ensuring nutrient requirements are met in formulating rations. The value assumed in the baseline analysis is that the crude protein requirement is met 95% of the time; that is, the hog producer is risk averse and is willing to accept nutrient requirement violations only 5% of the time. In contrast, if the producer is risk neutral to nutrient requirement violations then they would be willing to accept violations 50% of the time (Roush et al. 2007).
Fig. 2.
Fig. 2. (a) Distribution of cost savings from adoption of near-infrared reflectance spectroscopy (NIRS) (50% chance of meeting crude protein requirement, no premium penalty included) and (b) distribution of cost savings from adoption of NIRS (50% chance of meeting crude protein requirement premium penalties included). [Colour online.]
Fig. A1.
Fig. A1. Distribution of crude protein content (percentage) for barley wheat and peas. [Colour online.]
As illustrated in a comparison of Figs. 1 and 2 (i.e., 95% probability of non-violation versus 50% probability), a risk neutral farmer will benefit from the adoption of NIRS (i.e., positive cost savings) between 50% and 73% of the time, depending on whether premium penalties are considered in the calculations (Fig. 2); 498 of 1000 draws result in a benefit if no premium penalty is included, and 731 of 1000 draws lead to a benefit if a premium penalty is incorporated. In contrast, a risk averse farmer (Fig. 1) will benefit from NIRS adoption between 80% and 100% of the time. Because a risk neutral producer is not as concerned with nutrient requirement violations, and the penalties do not translate into significant premium losses, the benefits are not as great to the risk neutral producer as they are for the risk-averse producer.
Finally, the impact of the exchange rate is considered. The investment cost for the NIRS technology is converted from USD to CAD before being used in the NPV analysis. The exchange rate used to do this is $1.306CAD:$1USD.7 The impact of a change in the exchange rate would be to adjust the initial investment (i.e., INV in the NPV calculation) up or down, with a deterministic inverse corresponding impact on NPV values. A 10% increase in the exchange rate would result in initial investment requirements (expressed in CAD) of $35 912 and $158 014 for the sample and process analyzers, respectively, whereas a 10% decrease would result in corresponding investment requirements of $29 383 and $129 284, respectively.
The impact of the increased or decreased investment requirements would be decreased NPVs and increased NPVs, respectively, with the numerical effect being “dollar for dollar”. As with the sensitivity analysis for the discount rate, however, the qualitative results would remain unchanged.


The aim of the study is to estimate the cost benefit to Alberta hog producers of adopting NIRS machines and to determine if it is economically feasible for farmers to invest. A joint mathematical programming-simulation approach is used to estimate the cost savings generated by the use of NIRS machines. The results indicate mean savings of approximately $4 hog−1 yr−1 that can be attributed to adoption of NIRS technology (with the inclusion of premium penalties). For the 3453 head Alberta (i.e., small) hog farm this amounts to savings of over $13 600 yr−1 and for a larger 13 327 head farm savings are over $52 600 yr−1. These represent significant cost savings, but there are also significant investment and annual cost implications associated with adoption of NIRS technology, for both types of analyzers.
The empirical results presented here provide evidence to suggest that it is feasible and economically viable for larger hog farms to invest in the NIRS technology but that this is not necessarily the case for smaller-scale operators. Although the size of the farming operation affects the viability of this investment, the level of risk aversion of farmers also affects the potential benefit of investment in NIRS technology and will thus undoubtedly affect the likelihood of adoption. Specifically, the more risk averse the farmer, the greater the gains from investing in NIRS. Moreover, this study only considers the benefit from improving the accuracy of the protein content in a limited number of feedstuffs. The benefits would be undoubtedly larger if the protein and energy content in additional feedstuffs were assessed. NIRS has also been used to detect the presence of mycotoxins and as such, future work is directed at estimating the economic benefits of NIRS detection of mycotoxins in samples.
The question remains as to why there has been limited uptake of NIRS technology by hog producers in Alberta and more generally in western Canada. The results presented here certainly indicate that adoption by smaller producers is risky. Although more risk-averse producers may place greater value on having greater certainty in their knowledge of nutrient content, that risk aversion will also translate into greater requirements for return on investment and thus greater hesitation in adopting technology with uncertain benefits. For larger producers who purchase most of their feed, it also may be the case that they are less concerned with nutrient uncertainty due to either (a) guaranteed nutrient composition of pre-mixed feeds or (b) access to NIRS technology through their feed mill.8
Nevertheless, this study shows that there is economic value to the accurate estimation of basic nutrient levels within feedstuffs. In an industry where margins can be quite low, for some large producers, the investment in NIRS technology could be the difference between profits and losses, and long-term economic viability.


The authors thank Rob Hand, Mary-Lou Swift, Terry Veeman, Jason Wood, and John Folino for their assistance and advice on various aspects of this project. Funding in support of this project was provided by Alberta Crop Industry Development Fund.


Personal communication from Dr. Mary-Lou Swift, Director of NIRS/Technical Support at Trouw Nutrition.
Starch is used to predict the energy levels in feedstuff. In the data obtained from NRC (2012), swine energy targets as well as energy content of ingredients are readily available, but starch, ash, and fibre levels are not available for all ingredients of interest.
Information for each growth stage (initial and final weights, length of time, and nutrient requirements) is provided in the Appendix in Table A2.
The small farm size is based on the average size of hog operation in Alberta (Alberta Pork 2019), whereas the large farm size is based on a representative Manitoba farrow-finish farm (MAFRD 2018).
The actual sample analyzer modelled in this analysis uses near-infrared transmittance spectroscopy (NITS) technology. The NITS is similar to NIRS but operates under a narrower wavelength range. When calibrated, this allows for accurate testing of protein content.
Where expert opinion is noted through this paper, it refers to advice obtained from Dr. Mary Lou Swift who is an animal nutrition expert with Trouw Nutrition Canada.
This was the exchange rate effective 30 Dec. 2019, obtained from the Bank of Canada website.
As noted by an anonymous reviewer, this represents another NIRS scenario that could be modelled.


Alberta Agriculture and Forestry (AAF), Economics and Competitiveness Division, Statistics and Data Development Branch. 2019. Average farm input prices in Alberta. [Online]. Available from [29 Dec. 2019].
Alberta Pork. 2019. Annual report 2019. [Online]. Available from [20 Jan. 2020].
Campbell R.G., Taverner M.R., and Curic D.M. 1985. Effects of feeding level from 20 to 45 kg on the performance and carcass composition of pigs grown to 90 kg live weight. Anim. Prod. 40: 489–496.
Canada Pork International (CPI). 2013. Industry information: the Canadian pork story. [Online]. Available from [13 July 2019].
Cozzolino D. and Moron A. 2004. Exploring the use of near infrared reflectance spectroscopy (NIRS) to predict trace minerals in legumes. Anim. Feed Sci. Technol. 111: 161–173.
D’Alfonso J.A., Roush W.B., and Ventura T.H. 1992. Least cost poultry rations with nutrient variability: a comparison of linear programming with a margin of safety and stochastic programming models. Poult. Sci. 71: 255–262.
Manitoba Agriculture, Food and Rural Development (MAFRD). 2016. Guidelines for estimating swine farrow-finish costs 2016 in Manitoba. [Online]. Available from [13 July 2019].
Manitoba Agriculture, Food and Rural Development (MAFRD). 2018. Guidelines for estimating swine farrow-finish costs 2019-20 in Manitoba. [Online]. Available from [30 Dec. 2019].
Modroño S., Soldado A., Martínez-Fernández A., and de la Roza-Delgado B. 2017. Handheld NIRS sensors for routine compound feed quality control: Real time analysis and field monitoring. Talanta, 162: 597–603.
National Research Council (NRC). 1998. Nutrient requirements of swine, 10th rev. ed. National Academies Press, Washington, DC, USA.
National Research Council (NRC). 2012. Nutrient requirements of swine, 11th rev. ed. National Academies Press, Washington, DC, USA.
Norris K.H. 1996. History of NIR. J. Near Infrared Spectrosc. 4: 31–37.
Pomar C. and Remus A. 2019. Precision pig feeding: a breakthrough toward sustainability. Anim. Front. 9: 52–59.
Pujol A., Pérez-Vendrell A.M., and Torrallardona D. 2007. Evaluation of prediction of barley digestible nutrient content with near-infrared reflectance spectorscopy (NIRS). Livest. Sci. 109: 189–192.
Roush W.B., Purswell J., and Branton S. 2007. An adjustable nutrient margin of safety using linear and stochastic programming in an Excel spreadsheet. J. Appl. Poult. Res. 16: 514–520.
Western Hog Exchange. 2014. WHE: grading grids. Grids & Grading. Available from [14 July 2019].
Zhang H., Aalhus J.L., Gariépy C., Uttaro B., López-Campos O., Prieto N., et al. 2016. Effects of pork differentiation strategies in Canada on pig performance and carcass characteristics. Can. J. Anim. Sci. 96: 512–523.

Appendix A

Table A1.
Table A1. Feed ingredient prices.

Note: Source is MAFRD (2018), unless otherwise noted. DDGS, dried distillers grains with soluble.

This price was sourced from Alberta Agriculture and Forestry, Statistics and data development unit. AGDATA Series.
This price was sourced from [31 Dec. 2019].
This price was sourced from [31 Dec. 2019].
Table A2.
Table A2. Nutrient requirements for the seven hog growth stages.

Note: Requirements for all nutrients are minimum requirements, with the exception of feed intake, which represents the upper limit. kIU, kilo-international units; FTU, phytase international units; IU, international units. Sources are NRC (2012) and M.L. Swift, personal communication, Director of NIRS/Technical Support at Trouw Nutrition in Alberta, Canada.

Table A3.
Table A3. Distribution statistics for crude protein content (%), by ingredient.

Note: Source is M.L. Swift, personal communication, Director of NIRS/Technical Support at Trouw Nutrition in Alberta, Canada.

Table A4.
Table A4. Maximum allowed usage of each ration ingredient.

Note: For ration ingredients not included here (wheat, limestone, potassium chloride, di-calcium phosphate, and magnesium oxide), there are no restrictions on usage. NR, no restrictions; DDGS, dried distillers grains with solubles. Source is M.L. Swift, personal communication, Director of NIRS/Technical Support at Trouw Nutrition in Alberta, Canada.

Table A5.
Table A5. Relationship between crude protein (CP) and key amino acids.

Note: As noted in Table A4, there are restrictions on individual amino acid supplements allowed in the rations. To accurately model the level of amino acids provided in the ration, the relationships between amino acids and crude protein for key ingredients are incorporated into the ration models. Lys, lysine; Thr, threonine; Met, methionine; Cys, cysteine; Trp, tryptophan. Source is NRC (1998).

Table A6.
Table A6. Regression results estimating the marginal impact of crude protein (CP) deficiency.

Note: Based on data from Campbell et al. (1985). ANOVA, analysis of variance.

Table A7.
Table A7. Western Hog Exchange price premium grid. [Colour online.]

Information & Authors


Published In

cover image Canadian Journal of Animal Science
Canadian Journal of Animal Science
Volume 100Number 3September 2020
Pages: 557 - 569
Editor: Filippo Miglior


Received: 12 August 2019
Accepted: 22 February 2020
Accepted manuscript online: 25 March 2020
Version of record online: 25 March 2020


Request permissions for this article.

Key Words

  1. NIRS
  2. least cost ration models
  3. net present value
  4. technology adoption


  1. NIRS
  2. modèles de rations à moindre coût
  3. valeur actualisée nette
  4. adoption de technologie



Bijon A. Brown
Alberta Pork, Edmonton, AB T6E 5K1, Canada.
Henry An
Department of Resource Economics and Environmental Sociology, University of Alberta, Edmonton, AB T6G 2H1, Canada.
Scott R. Jeffrey [email protected]
Department of Resource Economics and Environmental Sociology, University of Alberta, Edmonton, AB T6G 2H1, Canada.


Copyright remains with the author(s) or their institution(s). Permission for reuse (free in most cases) can be obtained from

Metrics & Citations


Other Metrics


Cite As

Export Citations

If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download.

There are no citations for this item

View Options

View options


View PDF

Get Access

Login options

Check if you access through your login credentials or your institution to get full access on this article.


Click on the button below to subscribe to Canadian Journal of Animal Science

Purchase options

Purchase this article to get full access to it.

Restore your content access

Enter your email address to restore your content access:

Note: This functionality works only for purchases done as a guest. If you already have an account, log in to access the content to which you are entitled.





Share Options


Share the article link

Share on social media

Cookies Notification

We use cookies to improve your website experience. To learn about our use of cookies and how you can manage your cookie settings, please see our Cookie Policy.