Preventive Veterinary Medicine 114 (2014) 96–105

Contents lists available at ScienceDirect

Preventive Veterinary Medicine journal homepage: www.elsevier.com/locate/prevetmed

The Norwegian Healthier Goats programme – A financial cost–benefit analysis G. Elise Nagel-Alne a,e,∗ , Leif J. Asheim b , J. Brian Hardaker c , Liv Sølverød e , Dag Lindheim e , Paul S. Valle d,f a

Norwegian School of Veterinary Science, Department of Production Animal Clinical Science, P.O. Box 8146 Dep, N-0033 Oslo, Norway Norwegian Agricultural Economics Research Institute, P.O. Box 8024 Dep, N-0030 Oslo, Norway University of New England, Armidale, Australia d Norwegian School of Veterinary Science, Department of Food Safety and Infection Biology, P.O. Box 8146 Dep, N-0033 Oslo, Norway e TINE Extension Services, Goat Health Service, P.O. Box 58, N-1431 Ås, Norway f Kontali Analyse AS, Industriveien 18, 6517 Kristiansund, Norway b c

a r t i c l e

i n f o

Article history: Received 23 April 2013 Received in revised form 30 January 2014 Accepted 4 February 2014

Keywords: Caprine arthritis encephalitis Caseous lymphadenitis Johne’s disease Goat Eradication Stochastic simulation Cost–benefit

a b s t r a c t The aim of this study was to evaluate the profitability to dairy goat farmers of participating in the Healthier Goats disease control and eradication programme (HG), which was initiated in 2001 and is still running. HG includes the control and eradication of caprine arthritis encephalitis (CAE), caseous lymphadenitis (CLA) and paratuberculosis (Johne’s disease) in Norwegian goat herds. The profitability of participation was estimated in a financial cost–benefit analysis (CBA) using partial budgeting to quantify the economic consequences of infectious disease control through HG versus taking no action. Historical data were collected from 24 enrolled dairy goat herds and 21 herds not enrolled in HG, and supplemented with information from a questionnaire distributed to the same farmers. Expert opinions were collected to arrive at the best possible estimates. For some input parameters there were uncertainty due to imperfect knowledge, thus these parameters were modelled as PERT probability distributions and a stochastic simulation model was built. The CBA model was used to generate distributions of net present value (NPV) of farmers’ net cash flows for choosing to enroll versus not enrolling. This was done for three selected milk quota levels of 30 000 L, 50 000 L and 70 000 L, and both for before and after the introduction of a reduced milk price for the non-enrolled. The NPVs were calculated over time horizons of 5, 10 and 20 years using an inflation-adjusted discount rate of 2.8% per annum. The results show that participation in HG on average was profitable over a time horizon of 10 years or longer for quota levels of 50 000 L and 70 000 L, although not without risk of having a negative NPV. If farmers had to pay all the costs themselves, participation in HG would have been profitable only for a time horizon beyond 20 years. In 2012, a reduced milk price was introduced for farmers not enrolled in HG, changing the decision criteria for farmers, and thus, the CBA. When the analysis was altered to account for these changes, the expected NPV was positive over five years for the 50 000 L quota, indicating an increased profitability of enrolling in HG. The sensitivity analysis showed that particular attention should be paid to work load and investment costs when planning for disease control programmes in the future. © 2014 Elsevier B.V. All rights reserved.

∗ Corresponding author at: Ullevaalsveien 72, N-0454 Oslo, Norway. Tel.: +47 905 54 773; fax: +47 22 59 73 09. E-mail address: gunvor.elise.nagel [email protected] (G.E. Nagel-Alne). http://dx.doi.org/10.1016/j.prevetmed.2014.02.002 0167-5877/© 2014 Elsevier B.V. All rights reserved.

G.E. Nagel-Alne et al. / Preventive Veterinary Medicine 114 (2014) 96–105

1. Introduction In 2001, the Norwegian Goat Health Services initiated a programme to control and eradicate caprine arthritis encephalitis (CAE), caseous lymphadenitis (CLA) and paratuberculosis (Johne’s disease) in goat herds; the Healthier Goats programme (HG) (http://geithelse. tine.no/English). The programme was justified by the losses experienced from the prevalence of the three diseases in the Norwegian goat population. CAE is recognized as an important and costly disease of goats (Reina et al., 2009). Typically, CAE seropositive goats show a considerable reduction in milk yield compared to CAE seronegative goats, especially in later lactations (Martínez-Navalóna et al., 2013; Greenwood, 1995; Sanchez et al., 2001). The CAE virus (CAEV), a Lentivirus in the family Retroviridae, can cause a range of signs in infected goats, including arthritis, neurological dysfunction, progressive paresis and indurative mastitis. The most common presentation of the disease is arthritis. In 2004, bulk tank milk testing indicated that 88% of the milk producing herds in Norway had CAEV infection. CLA is a chronic infectious disease caused by Corynebacterium pseudotuberculosis, which mainly affects goats and sheep (Brown et al., 1987; Williamson, 2001; Baird and Fontaine, 2007). CLA, which occurs worldwide, is a potential zoonosis, and is especially evident in commercial flocks (Connor et al., 2000; Paton et al., 2003 and Peel et al., 1997). Holstad (1986) reported CLA in 19 out of 36 herds (53%) tested in a prevalence study in Northern Norway. Johne’s disease is caused by Mycobacterium avium subspecies paratuberculosis (MAP). In Norway, Johne’s disease has been more prevalent in goats than sheep and cattle (Djønne, 2003) and the disease has previously been endemic in goats in the southern part of the country. Differences in prevalence in different parts of the country are assumed to be related to climatic conditions, influencing the survival of the causative bacterium. By 1982, the infection rate of paratuberculosis in individual animals had been reduced from 53% to 1% following the introduction of a compulsory vaccination programme in 1967 (Saxegaard and Fodstad, 1985). Vaccination has largely reduced the prevalence of observed signs of the disease, but asymptomatic animals still shed MAP in their faeces (Djønne, 2003). Johne’s disease in goat herds is similar to that in cattle, with wasting and reduced milk yield (Juste and Perez, 2011; Cho et al., 2012; Mitchell et al., 2008), although goats do not exhibit diarrhoea. Historically, the main interest has been to eliminate CAE infection, owing to the serious effect on goat milk production (Nord et al., 1998). However, the basis for HG was the assumption that the snatching method (see explanation below) could enable the control and eradication of all three diseases, and thus that herds free of CAE, CLA and Johne’s disease could be established. Initially, the presence of CAE, CLA and Johne’s disease in enrolled herds was determined by diagnostic testing and clinical examination. was diagnosed from antibody detecCAE tion either in bulk tank milk or in individual serum samples using ELITEST-MVV/CAEV #CK104A

97

(http://www.mvdiagnostics.co.uk/documents/D750-02CK-104A-23-07-2008.pdf), in accordance with the test producers’ recommendations. When the bulk tank milk tests indicated infection with CAE, no further serological testing was conducted, whereas when the bulk tank milk tested negative for CAE, serological testing of individual animals was used to determine the prevalence of CAE in the herd. If the goats exhibited no clinical manifestation of CLA, serum samples were tested for CLA antibodies to determine the infection status of the herd using the ELITEST CLA #CK105A (http://www.mvdiagnostics.co.uk/ documents/CK105A02-08-07.pdf) in accordance with the test producers’ recommendations. Johne’s disease infection status of each herd was determined by the geographical location of the farm, as described above; hence no initial diagnostic tests were applied under HG. Herds with all three infections (i.e. CAE, CLA and Johne’s disease) were required to use the snatching method to eradicate the diseases. Herds that were free of CLA and Johne’s disease, and with less than 10% CAE seropositive goats, were allowed to choose the test-and-cull method, i.e. to cull the CAEpositive animals and retain the goats that tested negative. Until year 2009, 28 out of 199 herds enrolled in HG (14%) were allowed to apply this method. The snatching method involves the following four principles: (1) removal of the kids directly at delivery, (2) the does are not allowed to lick the kid, (3) kids are not allowed to suckle their mother, and (4) if labour is prolonged, the kid has to be isolated because of the higher risk of vertical transmission of infection. All snatched kids are then raised under controlled conditions at a separate location, and the old herd is eventually slaughtered at the end of lactation. Hygienic measures are implemented in farm buildings and nearby areas before the reintroduction of the snatched kids. HG has been financed jointly by the government, the dairy industry and the dairy goat farmers following a cost-sharing agreement by the partners. Governmental monetary support to HG has been determined as part of the annual national agricultural negotiations between farmer organizations and the Ministry of Agriculture and Food. Participation in HG has been, and still is, voluntary. However, from 2012, TINE Norwegian Dairies SA changed one major aspect by markedly lowering the price for goat milk from herds not enrolled in HG, thus creating a stronger economic incentive for non-enrolled farmers to participate. By August 2012, all dairy goat herds that deliver goat milk to TINE Norwegian Dairies SA had decided to join HG; effectively 100% of the Norwegian dairy goat population. In this largely retrospective financial cost–benefit analysis (CBA) we aimed to assess the net financial result of the programme at farm level by calculating the net present value (NPV) of enrolling in HG versus taking no action. Three different milk quota levels of 30 000 L, 50 000 L and 70 000 L were studied. We assessed the extra costs and revenues lost, as well as costs saved and revenues gained by farmers enrolling in the programme, compared to taking no action. We addressed both the initial programme

98

G.E. Nagel-Alne et al. / Preventive Veterinary Medicine 114 (2014) 96–105

Table 1 The number of herds entering the Healthier Goats programme (HG) from 1994 to 2013 using the snatching or test-and-cull method. Period

Number of herds in each category b

1994–2002a 2003–2006 2007–2011 2012–2013

Total

c

Snatching

Test-and-cull

6 75 155 116

9 14 11 20

15 89 166 136

352

54

406

a

Two herds undertook eradication in 1994 and 1996 before the initiation of HG in 2001. b Kids removed directly at delivery and raised in a clean compartment. c Infected animals are removed from the herd and production continues as before.

arrangements as well as the consequences of the change in milk pricing that occurred in 2012. 2. Materials and method 2.1. Data We were able to draw on two databases, both operated by the dairy goat industry: 1. The TINE Goat Milk Recording System (GMRS), which contains annual records of production and health parameters from farms. The participation rate for dairy goat producers was 87.7% in 2012. Herd-level data on milk yield, milk quality, herd size and milk quota were extracted from this database. 2. The TINE Efficiency Control Database for Goats (ECDG), which contains annual farm accounting data and is used mostly for economic consultations on member farms. The participation rate was 17% in 2008. From this database we gathered annually aggregated information about veterinary costs, milk revenues and milk price. Different ways of gathering relevant data were identified. Initially, the TINE Extension Services’ advisory personnel were asked to identify farms to include in the study sample; the reason for this approach was the local knowledge of these people about what information was available from different farms. We needed to draw the sample from farms with bookkeeping information available in ECDG. We also set a requirement of a herd size of more than 30 goats to include herds that were representative of the actual size distribution of Norwegian dairy goat herds. In 2009, 96% of all dairy goat herds had more than 30 goats (GMRS, annual reported data from 2009). Drawing on the available farm records in the ECDG, we were able to use information from 19 farms applying the snatching method and 5 farms applying the test-and-cull method. Likewise, we were able to collect information from bookkeeping records in the ECDG from 21 farms that had not enrolled in HG until after 2008, thus obtaining information on the consequences of taking no action. The numbers of farms in HG are presented in Table 1 according to the applied method for control and eradication.

The number of records of herd-level data in the two databases varied. Data from a minimum of two years before and two years after enrolling in HG were required, thus, GMRS and ECDG data from a minimum of five years were included in the calculations of NPV. We compared data from before and after enrolling in HG for the 24 enrolled farms and also compared these farms to the 21 farms not enrolled, for the same period (2002–2009). The latter comparison was carried out to assess the development that could have been expected without enrolling in HG. Quantitative data on farm effects of HG were collected with a questionnaire and included longevity of goats, feeding regimes, and the work-loads and investments in HG. We also collected qualitative data about the farmers’ experience or perceptions of disease prevalence before and after the eradication, and general effects of HG on the farm, including effects on the working environment and on the sustainability of the goat milk production. The questionnaire was developed in QuestBack® and was pre-tested by farmers not participating in the study and by veterinarians with extensive experience from dairy goat practice. The questionnaire was distributed to the 19 farmers applying the snatching method and to the five farms applying the test-and-cull method. Twenty-two of the farmers received a link to the web-based questionnaire by email, while the remaining two received a paper version of the questionnaire by mail. The information from the respondents was merged with GMRS and ECDG data, and contributed to the establishment of best estimates of budget parameters. We also drew on information in handbooks (NILF, 2010) to quantify feeding regimes, costs of raising goat kids and investment costs by farmers. Two earlier studies by Hardeng et al. (2009a,b), were used to assess the effects of HG on milk yield, milk quality and long-run replacement rates. These studies were based on milk records from the GMRS for individual goats from lactations 1, 2, 3 and =4, and provided estimates of the likely net changes in milk yield per year and lactation both in enrolled goat herds and in herds taking no action. Expert opinions were obtained from two experienced field personnel from TINE Extension Service. The experts were asked for their opinions when there was no other information available, e.g. for cost of feeding replacement kids and cost of feeding higher yielding does after enrolling in HG. Furthermore, the experts were asked for their assessment of input parameters gathered elsewhere, such as investment costs and work hours in relation to HG, as reported in the questionnaire. Although the sample size for the survey was small, we sought to triangulate parameter estimates using all available sources. 2.2. The financial cost–benefit analysis Drawing on all the sources noted above, farm-level estimates were made of the year-by-year costs and benefits of enrolling in HG and compared to not enrolling in HG. The NPV for the annual net cash flow (positive or negative) of enrolled farms was calculated. The year of investment was set to be 2009 and three different investment time horizons (5, 10 and 20 years) were studied. The model was

G.E. Nagel-Alne et al. / Preventive Veterinary Medicine 114 (2014) 96–105

applied to three different quota levels; 30 000 L (approx. 54 goats), 50 000 L (approx. 90 goats) and 70 000 L (approx. 125 goats), approximately corresponding to the 25, 50 and 75 percentiles of milk deliveries in 2009, respectively. The calculations were undertaken separately for the two eradication methods and the costs were weighted together to give a total programme assessment, using the 14% of farms that used the test-and-cull method up to 2009. Consequently, the resulting NPV reflects the average net benefit (positive, negative) at the farm-level across the two methods, i.e. the average population effect of HG for the given herd size or production level. This approach is relevant since the farmers do not know in advance what method they will have to apply. Furthermore, two scenarios were addressed; assuming that the test-and-cull method was applied by either all or none of the HG enrolled herds, respectively, i.e. each of the two methods were also studied separately. The introduction of a reduced milk price for nonenrolled farms markedly changed the economics of not participating in HG. Thus, the benefits of enrolling in HG differed between before and after 2012. The milk price for enrolled farms was NOK 8.28/L, whereas the milk price for non-enrolled farms was reduced by NOK 1.50/L (20%) after 2012. By adjusting the value of the goat milk for the nonenrolled farms, the NPV of participating in HG after 2012 was calculated. By the end of 2010, the total costs of the programme covered by the government and TINE BA had amounted to NOK 57.6 million. This included the costs of testing, laboratory analyses, administration and reimbursement to farmers. Costs for tests, analyses and administration amounted to NOK 1069 per goat, based on the total costs (NOK 25 191 000) divided by the 23 555 goats enrolled in HG by the end of 2010. The reimbursements constituted NOK 2300 per goat that were slaughtered as part of the programme. To assess the effect of this support, we added a scenario in which all the programme support from HG to farmers was assumed to have been discontinued. The NPVs of HG with and without compensation of these additional costs to farmers were calculated. 2.3. Stochastic parameters and stochastic dependency Several key parameters in the CBA were regarded as uncertain due to lack of full information (including no data about future values) and furthermore, there was variation across farms and time. We used stochastic estimates, in which the uncertainty was expressed as probability distributions to represent uncertain key parameters. We used PERT (Program Evaluation and Review Technique) distributions, a version of the ␤ distribution and recommended by Vose (2000) for such cases. A PERT distribution is defined by the minimum, most likely and maximum values, and these were assessed using available data combined with expert opinions. The assessed parameters for the uncertain input coefficients in the model are shown in Table 2. We used the concept of “total uncertainty”, advocated by Vose (2000), which includes both variation and uncertainty in the same distribution, i.e. a total uncertainty distribution. For parameters where it could have been

99

possible to separate variation from uncertainty (i.e. milk yield, feeding costs) data were not available to make this distinction. When two or more stochastic parameters are correlated, ignoring this stochastic dependency can lead to serious errors. In the cases where stochastic dependency was judged to apply, it was captured by incorporating the relevant correlation coefficients between those parameters in the model. Correlation matrices were specified for the dependencies between herd size and work hours/goat, between herd size and level of investment, and between milk yields from one lactation to the next. The Pearson correlation coefficients were calculated using the actual data. A sensitivity analysis to detect the impact of the different stochastic input parameters on NPV was performed by comparing the Pearson correlation coefficients between these inputs and the NPV. The higher the correlation between the input and NPV, the more significant the input is in determining the NPV. The calculations were conducted in Excel 2010 with the add-in @Risk® 5.7/6.0 (Palisade, NY) for stochastic analyses. Latin hypercube sampling was used in the simulations with 10 000 iterations, which is beyond the level needed to give stability in the simulated results (Further details of the calculations and the spreadsheet model will be provided upon request). 2.4. The costs of the HG programme Strict regulations were imposed under HG to ensure the success of the programme on each farm. At enrollment, there was a compulsory initial screening for the diseases, as described above, followed by an on-farm consultation to clarify necessary upgrading of buildings and nearby areas. Depending on the infections present, the appropriate eradication method was applied. Field veterinarians were employed and paid by the hour by HG to perform these tasks, thus farmers were not charged for the costs. Testing of animals for the three diseases was continued on a regular basis until five years after enrollment, again at no cost to the farmers. Farmers themselves had to handle the husbandry work load associated with the control and eradication process. They also had to invest in upgrading of buildings and in cleaning and disinfecting outdoor areas if this was deemed necessary by HG for the programme to succeed. 2.4.1. Investment costs of HG The investment cost of HG at the farm level (ICfarm level ), versus taking no action, was calculated as ICfarm level = W + I + R

(1)

where W = work load costs using the snatching method and denoted as: W=

2 

di ∗ w + (ni ∗ wn )

(2)

i=−2

i = 0 for year of enrolling HG, and di = day-time work hours in year i, ni = night-time work hours in year i, w = hourly wage valued as hired farm-work at NOK 148/h, wn = night-time hourly wage (wn = 1.25w). The extra

100

G.E. Nagel-Alne et al. / Preventive Veterinary Medicine 114 (2014) 96–105

Table 2 PERT (Program Evaluation and Review Technique) distributions with minimum, most-likely and maximum values, with their sources, of stochastic biological and economic input parameters of the model to estimate the farm level net present value (NPV) of participation in the Healthier Goats programme (HG) at farm level. Herd level parameters

Formula

Minimum

Most-likely

Maximum

Source

Investments, buildings (NOK) Other investments (NOK)a Year of enrolling HG, work daytime h/goat Year of enrolling HG, work nighttime h/goat 1 year after enrolling HG, work daytime h/goat 2 years after enrolling HG, work daytime h/goat Marginal work h/goat (not connected to HG)b

(1, 2) (1, 2) (1, 2) (1, 2) (1, 2) (1, 2) (6)

150 000 0 0.06 0.05 0.03 0.00 13.32

300 000 100 000 13.02 1.9 3.02 1.86 17.40

800 000 600 000 43.00 7.11 22.74 17.77 21.48

Norwegian goat milk price (NOK/L) after enrolling in HGc

(4, 5)

8.20

8.29

8.35

Concentrate price (NOK/kg) Roughage price (NOK/FEM) Extra feeding costs for kids after HG enrollment, %

(3, 6) (3, 6) (3)

3.00 2.80 5

3.15 3.50 20

3.30 4.20 25

Questionnaire Questionnaire Questionnaire Questionnaire Questionnaire Questionnaire NILF (2010) Stornes, O.K. (1990) TINE Norwegian Dairies SA Expert opinion Expert opinion Expert opinion

Milk yield: Confidence intervals (95%) and coefficient estimates for milk yield per lactation (kg milk from day 6 to day 275) in lactations 1, 2, 3 and ≥4 for the groups snatched and test-and culled before and after HG enrollment in period 1999–2000 and 2007–2008 and for the group non-enrolled are used to give minimum, most-likely and maximum values in the PERT distributions of milk yield used in Formula (4). a Any investments other than in buildings required by HG, e.g. fences on pasture or upgrading outdoor areas. b Stornes, O.K. (1990): Arbeidsforbruket i geitemelkproduksjonen (Title translation: Labour input in goat milk production). NILF, Oslo. c The goat milk price for the non-enrolled farms is assumed to be slightly lower than for the enrolled farms since the milk cell count number was higher in the non-enrolled hence attracting a price deduction. The milk price for the non-enrolled was calculated from the milk price of the enrolled subtracting the quality payment lost (0.28 NOK) due to higher milk cell count. We included a Bernoulli distribution for the non-enrolled milk price with a PERT distribution of (0.7, 0.9, 0.95) representing the “success” probability indicating the chance of a price reduction for the non-enrolled.

work load for snatching was estimated from data in the questionnaire. I = the sum of investments in buildings and other investments required by HG, and was derived from the questionnaire in which farmers were asked to only indicate the investments that were required by HG. R = the one-time cost of milk replacer used in raising snatched kids. From ICfarm level the net investment costs (NICfarm level ) was calculated as NICfarm level = ICfarm level + Pre ∗ Cre − Rprogram

(3)

where Pre = the probability of re-infection. The value for Pre (0.012) was based on information from 406 herds enrolled in HG from the beginning of the programme to the end of 2011. Five herds were re-infected, initiating a second round of eradication using the snatching method, with extra costs for both the farmers and the programme. We applied a Bernoulli distribution to Pre in which the probability of “success” = 0.012, as described above. The costs of re-infection for farmers, Cre , included the cost of cleaning and disinfecting the animal housing and pasture areas, and the costs of snatching and raising more kids. From the available data it was estimated that Cre = NOK 182 547 per re-infected farm. While re-infection could happen at any time after enrolling HG, in the calculations Cre was incorporated in the budget in the first year following the eradication process. Given the low probability of reinfection, this approximation was judged to be adequate for our purposes.

Rprogram = reimbursement to farmers by the programme.

2.4.2. Annual costs of HG The annual costs (Cann (i)) of HG at farm level in year i, versus taking no action, were calculated as Cann(i) = Fchange(i) + Rchange(i) + Vchange(i)

(4)

where Fchange(i) = the change in feeding costs in year i, Rchange(i) = change in replacement costs in year i, and Vchange(i) = change in veterinary costs in year i. The feeding costs associated with increased milk production were estimated based on an assumed requirement of 0.27 feeding units1 of concentrates and 0.20 feeding units of roughage per kg of additional milk produced. Concentrate and roughage feed prices were given PERT distributions (Table 2). The age composition of the new herd was assumed to have stabilized five years after applying the snatching method. Relatively more kids were raised to be kept as dairy goats during this five years’ adjustment period, adding to farmers’ costs. After five years, the long-run replacement rate was reduced to 25.4% compared to 32.2% in the non-enrolled herds. This was due to an increased longevity of the goats, assumed to be related to HG (Hardeng et al., 2009a,b). The amount and cost of milk replacer, concentrate and roughage needed for replacement kids were calculated. The feeding costs of rearing replacement kids were assumed to increase after HG and given a PERT distribution (Table 2). The total cost of raising each replacement kids was multiplied by the number of replacement kids needed in each year for five years following HG start up. The cost was markedly less for the test-and-cull method as most of the original herd was

1 Feeding unit (milk) = 6.9 MJ of energy, roughly equivalent to energy content of 1 kg of barley.

G.E. Nagel-Alne et al. / Preventive Veterinary Medicine 114 (2014) 96–105

retained, and the level of recruitment to the herd was therefore lower. The replacement costs for the non-enrolled herds were then subtracted to arrive at the net replacement costs due to HG. Based on the questionnaire responses, we estimated that the net annual veterinary cost increased by NOK 11 per goat in the years after enrolling in HG, and no other costs were incorporated. 2.5. Annual benefits of HG The benefits of HG to the dairy goat industry came from a reduced incidence and prevalence of the three targeted diseases. As a result, increased milk yields and improved milk quality after the eradication of CAE, CLA and Johne’s disease could be expected. The improved milk quality affected the milk price. The expected net change in milk revenue in year i (MRnet(i) ) was calculated as: MRnet(i) =

4+ 

(Yafter(j) ∗ AFafter(j,i) ∗ Psan

j=1

− Ybefore(j) ∗ AFbefore(j) ∗ Pbefore )

(5)

where Ybefore(j) and Yafter(j) = annual milk yield for goats in lactation j before and after HG, j = lactation number from 1 to 4+, AFafter(j,i) = fraction of goats in lactation j in year i after enrolling in HG, AFbefore(j) = fraction of goats in lactation j before HG enrollment, and Pbefore and Psan = the price of goat milk before and after enrolling in HG. We assumed no change in milk quota after HG enrollment; however, the quota can be produced with fewer animals due to a higher yield. 2.6. The net present value The net cash flow in the year of enrollment, Cflow(0) , comprises the net investment costs of HG (NICfarm level ) and is given a negative sign. For subsequent years, the net change, i.e. the difference between enrolling in HG and taking no action, in farm-level gross margin in year i (GMchange(i) ) was calculated as GMchange(i) = MRnet(i) − Cann(i)

(6)

From (6) we calculated the net cash flow in year i after HG enrollment, Cflow(i) as Cflow(i) = GMchange(i) − Sloss(i) + Vwork(i) + Vfeed(i)

(7)

where Sloss(i) = lost agricultural subsidy payment because of fewer animals in the herd after enrolling HG (subsidies are given per animal) in year i, Vwork(i) = the value of saved work time because of a reduced herd size in year i, and Vfeed(i) = the value of saved maintenance feed because of reduced herd size in year i. Agricultural subsidies amounted to NOK 2001 per goat in 2009. The saved work time was estimated from the marginal work time per goat, valued as hired labour and given a PERT distribution (Table 2). Maintenance feed per goat was 229

101

feeding units of roughage and 55 feeding units concentrate (NILF, 2010). For the deterministic case, with stochastic variables set at their expected values, the NPV of HG was calculated as: NPV =

n  Cflow(i) i=0

(1 − r)i

(8)

where i = year, n = the time horizon, Cflow(i) = the net cash flow in year i, and r = the discount rate. The basis for setting the discount rate, r, was an average 10-year bond rate of 4.3% per annum adjusted by an average inflation rate of 1.5% for the three years 2010–2012, resulting in an approximate discount rate of 2.8% p.a. An inflation adjusted rate was appropriate because the NPV budget was calculated in constant 2009 prices. The net financial results of implementing HG were derived as cumulative distribution functions (CDFs) of NPV which represent the present values of the uncertain annual net cash flows to the chosen time horizons. The CDFs of NPV can be used as a basis for choice accounting for the client farmers’ individual attitudes to risk. Without knowledge of the applicable degrees of risk aversion, the CDFs allow the probability of a negative NPV to be identified, providing a useful measure of riskiness. 3. Results 3.1.1. The NPV of HG with reimbursement to farmers Representing the situation before the introduction of a reduced milk price for the non-enrolled farms, the model was applied to the three different quota levels with three different time horizons; 5, 10 and 20 years. An example of the results generated is presented in Fig. 1. For each quota level, the CDF shows, on the vertical axis, the probability of getting a value of NPV less than any corresponding point on the horizontal axis. The case shown is for a quota level of 50 000 L. A broadly similar pattern of results was generated at other quota levels. Fig. 1 shows how the CDFs are shifted rightwards as the time horizon is extended. This is the result of including more future years of benefit from the elimination of diseases in the NPV calculations. The figure also shows the intersection of the CDFs with the vertical line drawn at the origin on the NPV scale, indicating the probabilities of a negative NPV. These probabilities, as recorded in the figure, are 0.919 for a 5-year horizon, 0.249 for 10-year horizon and 0.011 for a 20-year horizon. Results for all comparable cases are summarized in Table 3. Table 3 shows that expected NPVs were found to be negative for all quota levels with a time horizon of five years. Evidently five years was too short a time period for farmers to recoup their investment costs. However, for horizons of 10 and 20 years, the calculated expected NPVs were found to be positive for quota levels of 50 000 L and 70 000 L and only slightly negative at 30 000 L. Over a 10-year horizon with a quota of 30 000 L, the probability of a negative NPV was 55%. At quota levels of 50 000 L and 70 000 L the chance of having a negative NPV was reduced to 25% and 15%, respectively. Our results indicate that, over 20 years, there

102

G.E. Nagel-Alne et al. / Preventive Veterinary Medicine 114 (2014) 96–105

Fig. 1. Cumulative distribution of the net present value (NPV) representing the situation before the introduction of a reduced milk price for farmers not enrolled in the Healthier Goats programme, over three different time horizons (5, 10, 20 years) with a milk quota of 50 000 L. Figures in % give the proportions of expected values of NPV below/above 0.

was still a 5% chance of experiencing a negative NPV with a quota of 30 000 L. However, for the other quota alternatives the chances were only 1% and 0.6%. Responses from the questionnaire were provided by 17 farmers using the snatching method and by two farmers using the test-and-cull method, altogether 19 respondents, giving an overall response rate of 79%. By the end of 2010, the total costs of the programme (covered by the government and TINE) amounted to approximately NOK 60 000 000, covering reimbursements, testing costs and administration. However, fewer animals in the herds after the eradication of CAE, CLA and Johne’s disease meant a reduction in subsidy payments to the farmers of approximately NOK 50 000 per year at a quota of 50 000 L (median quota level). By the end of 2010, 255 dairy goat herds were enrolled in HG, with an estimated annual total of NOK 12 750 000 in subsidies saved by the government.

Table 3 Summary of results from @Risk® simulation to compare the effect of three different milk quota levels (30 000 L, 50 000 L, 70 000 L) on the expected net present value (NPV) over three different time horizons (5, 10, 20 years) representing the situation before the introduction of a reduced milk price for farmers not enrolled into the Healthier Goats programme. The expected NPV are given in NOK 1000. Quota (L)

5-year horizon Expected value of NPV Probability of NPV < 0 NOK 10-year horizon Expected value of NPV Probability of NPV < 0 20-year horizon Expected value of NPV Probability of NPV < 0

30 000 L

50 000 L

70 000 L

−208 98%

−196 92%

−185 84%

−15 55%

124 25%

263 15%

297 5%

646 1%

994 0.6%

3.1.2. Model evaluation, sensitivity analysis Table 4 shows that with a quota of 30 000 L, the three input parameters (i) work hours in year of enrolling in HG, (ii) investments in buildings, and (iii) other investments in relation to HG had the highest absolute (negative) correlation coefficients for all three time horizons. The milk yield in the fourth and subsequent lactations after applying the snatching method had the highest positive correlation coefficient (0.20) for the 30 000 L quota over 20 years. This illustrates that the increased milk yield strongly affected the NPV over the long-time horizon, although not as strongly as the impacts of some of the cost elements. With a quota of 50 000 L over 5 and 10 years, the situation was found to be similar to the 30 000 L case; that is, the same three parameters had the highest absolute (negative) correlation coefficients. A different situation emerged over 20 years. Here, the work hours in the year of HG enrollment still had the highest absolute (negative) correlation coefficient (−0.32), but second, with a positive correlation coefficient (0.21), came the effect of milk yield in the fourth and subsequent lactations after enrolling for snatching. Third came the marginal work input, i.e., the level of work hours on the farm prior to HG, with a positive correlation coefficient (0.17). The milk yields in the fourth and subsequent lactations before enrolling in HG (−0.17) was equally important for the result over 20 years for the snatching group. A similar situation was revealed with a quota of 70 000 L over a 20-year time horizon. The correlation coefficients with the highest relative influence was the work hours in the year of HG enrollment (−0.32), then the effect of milk yield in higher lactations after enrolling for snatching (0.22), followed by the marginal work input (0.18). Even over a 20-year time horizon with the highest quota level of 70 000 L; the parameters representing the investment costs had clearly the highest impact on the resulting NPV, but the influence of the increased milk yield and the work time per

G.E. Nagel-Alne et al. / Preventive Veterinary Medicine 114 (2014) 96–105

103

Table 4 Sensitivity analysis of stochastic input parameters presented by correlation coefficients, indicating the level of influence on the expected NPVs for three quota levels (30 000 L, 50 000 L, 70 000 L) at time horizons of 5, 10 and 20 years. Stochastic input parameter

30 000 L 5y

10 y

20 y

5y

10 y

20 y

5y

10 y

20 y

Year of HG enrollment, work time day h/goat Investments, buildings (NOK) Other investments (NOK) 1 year after HG enrollment, work time day h/goat 2 years after HG enrollment, work time day h/goat day/goat Milk yield 2007–08, 1.lactation after HG enrollment, by snatchinga Milk yield 2007–08, ≥4.lactation after HG enrollment, by snatchinga Marginal work time h/goat, hoursb Year of HG enrollment, work time night h/goat Milk yield 2007–08, 2.lactation after HG enrollment, by snatchinga Milk yield 1999–00, ≥4.lactation before HG enrollment, by snatchinga

−0.52 −0.49 −0.43 −0.25 −0.19 0.11 – 0.10 −0.09 0.09 −0.09

−0.43 −0.39 −0.35 −0.20 −0.16 0.12 0.13 0.14 – 0.11 −0.14

−0.31 −0.27 −0.24 −0.15 −0.11 0.11 0.20 0.17 – – −0.17

−0.62 −0.34 −0.34 −0.30 −0.23 0.13 – 0.11 −0.11 0.11 −0.10

−0.47 −0.26 −0.23 −0.22 −0.17 0.13 0.15 0.16 – 0.12 −0.15

−0.32 −0.17 −0.15 −0.15 −0.12 0.12 0.21 0.17 – – −0.17

−0.65 −0.26 −0.23 −0.31 −0.24 0.13 – 0.12 −0.12 0.11 −0.10

−0.48 −0.19 −0.17 −0.23 −0.18 0.13 0.15 0.16 – 0.12 −0.16

−0.32 −0.13 – −0.16 −0.12 0.12 0.22 0.18 – 0.12 −0.17

50 000 L

70 000 L

a Explanation of milk yield parameters: Milk yield 2007–08, 1.lactation after HG enrollment, by snatching: Milk yield measures in years 2007–08 after HG enrollment in first lactation goats; Milk yield 2007–08, 2.lactation after HG enrollment, by snatching: Milk yield measures in years 2007–08 after HG enrollment in second lactation goats; Milk yield 2007–08, ≥4.lactation after HG enrollment, by snatching: Milk yield measures in years 2007–08 after HG enrollment in fourth and higher lactation goats; Milk yield 1999–00, ≥4.lactation before HG enrollment, by snatching: Milk yield measures in years 1999–00 before HG enrollment in fourth and higher lactation goats. b See reference footnote (b) in Table 2.

goat showed higher influences on the resulting NPV as the time horizon was extended. The stochastic input parameters in Table 4 are those that were detected as having significant impacts on the expected NPV in the sensitivity analysis. The effect of prices of concentrate, roughage and cost of feeding replacement kids are not reported, as these inputs had very low influence on the expected NPVs. 3.2. The NPV without reimbursement to farmers When reimbursement to farmers was subtracted from the benefits, the expected NPV became negative for all three quota levels for 5- and 10-year time horizons. However, over 20 years the expected NPV was still positive (Table 5). The CDFs of the NPVs showed that there is a 100% probability of a negative NPV over a time horizon of five years in this situation, independent of the quota level. The expected NPV over five years was lower for the 50 000 L Table 5 Summary of results from @Risk® simulation showing the effect of no compensation of costs on the expected net present value (NPV) for the situation before the introduction of a reduced milk price for farmers not enrolled in the Healthier Goats programme, for the three different quota limits (30 000 L, 50 000 L, 70 000 L) over 5, 10, 20 year time horizon. Expected values of NPV are given in NOK 1000. Quota (L)

5-year horizon Expected value of NPV Probability of NPV < 0 10-year horizon Expected value of NPV Probability of NPV < 0 20-year horizon Expected value of NPV Probability of NPV < 0

30 000 L

50 000 L

70 000 L

−378 100%

−479 100%

−581 100%

−185 94%

−158 81%

−132 70%

128 23%

364 9%

598 6%

and 70 000 L quotas compared to the 30 000 L quota, presumably because losses are greater at a larger scale of production, just as the gains are greater when costs are reimbursed. At the highest quota level examined, the chance of a negative NPV was 6% for a 20-year time horizon without reimbursement.

3.3. Consequence of reduced milk price for non-enrolled farms from 2012 In the basic model we assumed only a slight increase in the price for milk of improved quality following lower somatic cell counts after the eradication. A separate scenario was developed to examine the economic effects of the reduced milk price imposed on non-enrolled farms in 2012. By conducting the analysis with the reduced milk price for the decision of taking no action, the expected NPVs changed considerably. The expected NPV for the 50 000 L quota with a 5 years’ time horizon was 147 000 NOK, with a 15% chance of having a negative NPV. As expected, the NPVs for the other quota levels and time horizons showed a similar pattern, with a higher expected NPV and lower chances of a negative NPV compared to the situation before the milk price reduction.

3.4. Test-and-cull versus snatching method, and their impact on the expected NPV Separate calculations of the expected NPV of the two different eradication methods; snatching and test-and-cull, showed that the NPV of the ‘all use’ test-and-cull method for the 50 000 L quota at a 10 years’ time horizon was 370 000 NOK, compared to 80 000 NOK for ‘none use’ testand-cull method with the same quota and time horizon. The higher cost of the snatching method is clear. However, NPV is still positive and the snatching method should be used when necessary.

104

G.E. Nagel-Alne et al. / Preventive Veterinary Medicine 114 (2014) 96–105

4. Discussion

4.3. External validation

4.1. Distributions of NPV in relation to scale of production and time horizon

The non-enrolled farms represented the general situation among the dairy goat farms without eradication of CAE, CLA and Johne’s disease. They represented the same as the data from the enrolled farms prior to enrolling in HG (own controls). The study by Hardeng et al. (2009a) showed that the average milk yield was higher for the nonenrolled group than for the enrolled group prior to HG. This implies that the loss in milk yield for the enrolled farms might well have been even more substantial without HG. Hence, the difference between enrolled and nonenrolled, as described in our CBA model, could have been underestimated, suggesting that the CBA model provided a somewhat conservative assessment of the benefits of HG. The overall detrimental effect of CAE, CLA and Johne’s disease on the productivity of dairy goats was difficult to measure or value. It was not possible to distinguish between the impacts of the individual diseases. With this in mind, caution is needed when comparing HG to disease control programmes targeting only one or two of the diseases, as the costs of HG were spread across the elimination of three diseases. However, with this reservation, we believe that HG is a relevant model for other disease control programmes, either in Norway or other countries, just as the Norwegian BVDV control and eradication programme (Valle et al., 2005) was an important reference in the establishment of HG.

For the situation before the price reduction was introduced for the non-enrolled farms, our results showed that, over a 5-year time horizon, the high investments required for the farmers’ participation in HG resulted in a negative expected NPV, regardless of the quota levels. With large investments at the start of the programme, the farmers faced a high risk of a net loss over the first few years. On the other hand, over 10- and 20-year horizons, expected NPVs were positive both at 50 000 L and 70 000 L quota levels. Yet over a 10-year period there were significant risks of farmers making a loss at all quota levels, but especially at the lowest scale of production. Evidently, control and eradication of CAE, CLA and Johne’s disease required farmers to take a long-term view – as in the short term they faced significant downside risks. Those risks were most serious for farmers with small herds, who may be the least able to take a long-term view and to bear the risk of the investments. The discount rate was held constant in the modelling and reflected the situation with a low and stable interest rate, which Norway has experienced over several years. This has clearly contributed to make the programme more profitable and worthwhile undertaking. A higher discount rate, such as might be appropriate for a farmer with substantial existing debts, would make the programme less profitable in expected value terms, and more risky. Whether a goat farmer would take a short- or a longterm view of the investment may depend on personal circumstances, such as age or level of indebtedness and the upcoming capital needs of the farm and family. These factors represent the overall objectives a farmer has with his/her business. We did not study the farmers’ decisions criteria, but we believe an insight into these aspects would be useful when addressing disease control issues in the future.

4.2. Effects of government reimbursement In the scenario in which farmers were left without any reimbursement of the costs of control and eradication, and also had to pay for testing and analyses themselves, the expected NPV was negative over 10 years, independent of the quota level. Thus, the importance of governmental support to the programme to make it financially attractive was demonstrated. Whether the prolonged period before gaining a positive NPV in the scenario without reimbursements would have affected farmers’ decisions to enroll in the programme, has not been assessed. It seems likely that more farmers, if not most, would have been discouraged by the extended time horizon before the programme would be financially beneficial. Certainly, industry representatives claim that the governmental support has been crucial for the success. The farmers’ assessment in this respect has not been studied.

4.4. Additional benefits not estimated in the model By initiating and implementing HG, a major disease burden has been removed from the Norwegian goat population and consequently animal welfare has been improved (Muri et al., submitted for publication). Because such benefits are hard to quantify in terms of monetary value, they are difficult to include in an economic comparison (Dijkhuizen and Morris, 1997) and therefore not captured in the present NPV calculations. However, the satisfaction of having a healthy herd with improved animal welfare and a reduction in potential human health risks are important factors when considering the full benefits of HG. In 2012, the main focus of HG shifted towards recruiting the remaining farms, in order to complete the eradication in the entire Norwegian dairy goat population. Achieving that goal would lower the risk of re-infection of herds free from CAE, CLA and Johne’s disease. The reduction in milk price was implemented to ensure enrollment of the total Norwegian dairy goat population in HG, and made nonenrollment an unsustainable alternative for the majority of the dairy goat farmers. The extra benefit of eventually obtaining a close to zero risk of re-infection, given all herds are cleared of the diseases, was not explicitly included in our analyses. The subsidies saved by the government can be seen in relation to the governmental monetary support to HG, as the support to HG is a one-time payment per farm, and following this, a continuous reduced subsidy payment can be achieved due to higher yielding animals. We observe a theoretical opportunity for the government to recover the support to HG in less than 5 years, given that the

G.E. Nagel-Alne et al. / Preventive Veterinary Medicine 114 (2014) 96–105

quota system and regulations on subsidy payments remain unchanged. Hence, from a government point of view, it can be argued that providing money to HG could be seen as a good investment. We did not undertake a social economic cost–benefit analysis of the programme from the viewpoint of society as a whole. That is a matter for further research. 5. Conclusion The Norwegian Healthier Goats programme was shown to be profitable at farm level over 10 years and more, as measured by the positive expected NPV of investments. Nevertheless, we found a significant risk of making loss over a 10-year time horizon for farmers with quota levels around 50 000 L. Moreover, there was a higher chance that some smaller farms would have invested in the programme unprofitably over reasonably long time horizons. Given the relatively small herd sizes in the Norwegian dairy goat industry, the smallest herds’ apparent economic disincentives for enrollment could become a source of re-infection for the population as a whole, implying that specific and/or differentiated measures should be considered. There are lessons to be learned from identifying the main factors that influenced the profitability of CAE, CLA and Johne’s disease eradication. Factors we considered included the effect of farm size, the level of investment that could be justified by programme gains, and the impact of programme support. The results of these analyses, although obviously specific to the Norwegian dairy goat population, may be useful in evaluating or planning similar schemes in the future for other dairy goat populations, or even other farm animals. To our knowledge, there are presently no other financial evaluations of disease control programmes for CAE, CLA and Johne’s disease in goats. Conflict of interest statement All authors have disclosed any financial and personal relationships with other people and organizations that could inappropriately have influenced our work on this paper. Acknowledgements The first author is grateful for the received financial support from TINE Norwegian Dairies SA, the Norwegian School of Veterinary Science, Department of Production Animal Clinical Science, and the Norwegian Research Council (Project number: 179745). We wish to thank all goat farmers for contributing to our study. References Baird, G.J., Fontaine, M.C., 2007. Corynebacterium pseudotuberculosis and its role in ovine caseous lymphadenitis. J. Comp. Pathol. 137, 179–210.

105

Brown, C.C., Olander, H.J., Alves, S.F., 1987. Synergistic hemolysisinhibition titers associated with caseous lymphadenitis in a slaughterhouse survey of goats and sheep in Northeastern Brazil. Can. J. Vet. Res. 51, 46–49. Cho, J., Tauer, L.W., Schukken, Y.H., Gómez, M.I., Smith, R.L., Lu, Z., Grohn, Y.T., 2012. Economic analysis of Mycobacterium avium subspecies paratuberculosis vaccines in dairy herds. J. Dairy Sci. 95, 1855–1872. Connor, K.M., Quirie, M.M., Baird, G., 2000. Characterization of United Kingdom isolates of Corynebacterium pseudotuberculosis using pulsedgel electrophoresis. J. Clin. Microbiol. 38, 2633–2637. Dijkhuizen, A.A., Morris, R.S., 1997. Animal Health Economics. Principles and applications. University of Sydney. Djønne, B., 2003. Paratuberculosis on goats – a special focus on the Nordic countries. Acta Vet. Scand. 44, 257–259. Greenwood, P.L., 1995. Effects of caprine arthritis-encephalitis virus on productivity and health of dairy goats in New South Wales, Australia. Prev. Vet. Med. 22, 71–87. Hardeng, F., Sølverød, L., Leine, N., Valle, P.S., Lindheim, D., Nagel-Alne, G.E., Østerås, O., 2009a. Mjølkeytelse og holdbarhet i sjukdomssanerte geitebuskaper. In: Fog, M.O. (Ed.), Proceedings Husdyrforsøksmøtet 2009 (Production Animal Science Meeting 2009). Lillestrøm, Norway, 11–12 February, pp. 287–290. Hardeng, F., Sølverød, L., Valle, P.S., Nagel-Alne, G.E., Østerås, O., 2009b. Celletall i geitmelk, effekt av saneringsprogrammet “Friskere Geiter”. In: Fog, M.O. (Ed.), Proceedings Husdyrforsøksmøtet 2009 (Production Animal Science Meeting 2009). Lillestrøm, Norway, 11–12 February, pp. 291–294. Holstad, G., 1986. Corynebacterium pseudotuberculosis infection in goats II. The prevalence of caseous lymphadenitis in 36 goat herds in northern Norway. Acta Vet. Scand. 27, 584–597. Juste, R.A., Perez, V., 2011. Control of paratuberculosis in sheep and goats. Vet. Clin. North Am.: Food Anim. Pract. 27, 127–138. Martínez-Navalóna, B., Perisb, C., Gómezc, E.A., Perisd, B., Roched, M.L., Caballerod, C., Goyenae, E., Berriatua, E., 2013. Quantitative estimation of the impact of caprine arthritis encephalitis virus infection on milk production by dairy goats. Vet. J. 197 (2), 311–317. Mitchell, R.M., Whitlock, R.H., Stehman, S.M., Benedictus, A., Chapagain, P.P., Grohn, Y.T., Schukken, Y.H., 2008. Simulation modeling to evaluate the persistence of Mycobacterium avium subsp. paratuberculosis (MAP) on commercial dairy farms in the United States. Prev. Vet. Med. 83, 360–380. Muri, K., Valle, P.S., Leine, N., 2014. Welfare effects of an eradication programme against caprine arthritis encephalitis, caseous lymphadenitis and paratuberculosis in Norwegian dairy goat herds. J. Livest. Sci. (submitted for publication). NILF, 2010. Handbook of farm planning 2009/2010. The Norwegian Agricultural Economics Research Institute, Oslo. Nord, K., Løken, T., Orten, Å., 1998. Control of caprine arthritis-encephalitis virus infection in three Norwegian goat herds. Small Rumin. Res. 28, 109–114. Paton, M.W., Walker, S.B., Rose, I.R., 2003. Prevalence of caseous lymphadenitis and usage of caseous lymphadenitis vaccines in sheep flocks. Aust. Vet. J. 81, 91–95. Peel, M.M., Palmer, G.G., Stacpoole, A.M., Kerr, T.G., 1997. Human lymphadenitis due to Corynebacterium pseudotuberculosis; report of ten cases from Australia and review. Clin. Infect. Dis. 24, 185–191. Reina, R., Berriatuac, E., Lujánd, L., Justee, R., Sánchez, A., Andrésa, D., de Amorena, B., 2009. Prevention strategies against small ruminant lentiviruses: an update. Vet. J. 182, 31–37. Sanchez, A., Contreras, A., Corrales, J.C., Marco, J.C., 2001. Relationships between infection with caprine arthritis encephalitis virus, intramammary bacterial infection and somatic cell counts in dairy goats. Vet. Rec. 148, 711–714. Saxegaard, F., Fodstad, F.H., 1985. Control of paratuberculosis (Johne’ diseases) in goats by vaccination. Vet. Rec. 116, 439–441. Valle, P.S., Skjerve, E., Martin, W., Larssen, R.B., Østerås, O., Nyberg, O., 2005. Ten years of bovine virus diarrhoea virus (BVDV) control in Norway: a cost-benefit analysis, 2005. Prev. Vet. Med. 72, 189–207. Vose, D., 2000. Risk Analysis – A quantitative Guide, second ed. John Wiley, Chichester. Williamson, L.H., 2001. Caseous lymphadenitis in small ruminants. Vet. Clin. North Am.: Food Anim. Pract. 17, 359–371, vii.