The potential impact of imposing best management practices for nutrient management on the US broiler industry

Journal of Environmental Management (2000) 60, 145–154 doi:10.1006/jema.2000.0378, available online at http://www.idealibrary.com on

The potential impact of imposing best management practices for nutrient management on the US broiler industry C. S. McIntosh* , T. A. Park and C. Karnum

The imposition of nutrient management plans for disposal of poultry litter will increase broiler production costs. This research examines the potential impacts of these cost increasing events on the US broiler industry. The results show that for 8, 40 and 80% increases in costs, wholesale prices eventually return to previous levels, and production levels stabilize at slightly lower levels.  2000 Academic Press

Introduction The US broiler industry has seen far reaching structural changes in the last 50 years. Extensive improvement in technology and integration of various levels of production, processing and marketing have transformed the broiler industry from a disorganized group of small, independent farms, processors and distributors to a highly integrated, concentrated and efficient operation. Advances in breeding, nutrition, housing, equipment and disease control have permitted large-scale production of high quality meat and reduced real costs of production. Vertical integration has resulted in substantial investment in housing and equipment while reducing operating expenses. These factors have led to explosive growth in the number of broilers produced and increased geographical concentration of production. This growth in, and concentration of, poultry production has not come without negative impacts. One of the most critical concerns facing poultry growers today is how to dispose of the waste products from production. The two primary waste products from poultry production are poultry litter (a combination of poultry feed, manure, and bedding materials) and dead birds. Here, we will focus on the economic and environmental impacts from the disposal of poultry litter. 0301–4797/00/100145C10 $35.00/0

At present, most poultry litter is land applied, and its application is largely unregulated. There is, therefore, potential for contamination of ground and surface water in areas near the land application sites (Madison and Brunett, 1985; Council for Agriculture Science and Technology, 1996; Edwards et al., 1996; Patterson, 1996; Bosch et al., 1997). These concerns have, in some cases, led to the imposition of state-level nutrient management regulations on land application of manures (Carson and Smeltz, 1993). Environmental concerns have also led to processors requiring growers to consider and implement best management practices (BMP) for land application of litter. A BMP is ‘a practice or combination of practices that is determined by the state (or designated area-wide planning agency), after problem assessment, examination of alternative practices, and appropriate public participation, to be the most effective practicable (including technological, economic, and institutional considerations) means of preventing or reducing the amount of pollution generated by nonpoint sources to a level acceptable with water quality goals’ (Bailey and Waddell, 1979). By definition, BMPs have the potential for reducing the water quality impacts from agricultural production. However, it is quite difficult to estimate a priori the effectiveness of a particular BMP when that BMP is applied under different conditions than those

Ł Corresponding author Agriculture, Economic and Rural Sociology Department University of Idaho, Idaho Falls, IO 83402–1575, USA Received 15 October 1997; accepted 9 August 2000  2000 Academic Press

146

C. S. McIntosh et al.

under which it was developed (Edwards et al., 1996). The major BMPs associated with poultry litter disposal are nutrient management, waste utilization and pasture and hayland management. Nutrient management is defined as ‘managing the amount, form, source, placement, and timing of applications of plant nutrients’ (Soil Conservation Service (SCS), 1992). Waste utilization is defined as ‘using agricultural waste or other waste on land in an environmentally acceptable manner while maintaining or improving soil and plant resources’ (SCS, 1987a). Pasture and hayland management consists of ‘proper treatment and use of pasture and hayland’ (SCS, 1987b) and includes guidelines for grazing times, weed control, and forage harvest. These three BMPs are not mutually exclusive and should be considered in concert. Large, integrated poultry companies may ‘strongly recommend’ and in some cases require, as part of the growers contract, a series of BMPs for handling and disposal of poultry litter. Such companies may also claim that the responsibility for protecting surface and groundwater lies with the individual growers (Gerstenzang, 1997). Bearing the responsibility for controlling non-point source pollution will undoubtedly raise the grower’s costs of production. It is the potential industry-wide impacts of these increased costs that are the subject of this research. Though little direct evidence of the costs directly associated with poultry litter BMPs appears in the literature, Clouser et al. (1994) have identified cost increases associated with nutrient management plans for dairy farms in Florida. Their analysis indicates that when BMPs were in place, variable production costs were increased by eight percent. Bosch and Pease (1991) examined combinations of dairy and poultry production and found that when land applications were restricted by nutrient management plans to the lesser of the limiting nutrient (N or P) net returns to livestock production were decreased 38 to 142%. Moore et al. (1995) found that the amount of land required for manure application was five to 10 times greater when manure management was based on P vs. N. Therefore, the impact to growers of BMP imposition also depends upon which

nutrient is used as a basis for determining land application levels. The purpose of this paper is to examine the potential impacts on a national basis to the US poultry industry resulting from the imposition of nutrient management regulations (enforcement of BMPs individually or in combination). Potential regulation of the disposal of poultry wastes will have significant impacts on cost and competitiveness in the industry, potentially increasing the cost of food on supermarket shelves, while making US farming less competitive (Warrik and Goodman, 1998). Such regulations will increase poultry production costs by increasing the amount of labor, management, and land needed to dispose of litter (Bosch et al., 1997). These costs have been estimated at as much as US$3500 per farm and from 5–103% of total gross sales per farm (Westenberg and Letson, 1995). States have adopted a variety of regulatory policies and incentives that impact manure management systems in the poultry and livestock industry. Some states have adopted mandatory regulations involving permits or approved manure management systems or nutrient management plans for poultry and livestock operations (for example, North Carolina Leavenworth, 1996). Other states are encouraging the adoption of specific BMP while others specify only minimal design criteria (for example, Arkansas, Edwards et al., 1996). The CAST (1996) report on integrated waste management noted that these regulations are difficult to enforce and often ignored. Producers view the varying regulations at the state level as expensive and arbitrary, and they are resistant to implementing these plans. Given the enforcement difficulties and resentment toward state level plans, it is important to develop industry level models to assess the impact of manure management strategies on industry production costs and the dynamic adjustment of production, prices and consumption. In order to examine how nutrient management regulations may impact the industry, we examine the dynamic interrelationships within the industry over time. This can be accomplished using a multivariate timeseries model. Multivariate time-series models, such as the vector autoregressive (VAR) model used here, are designed to examine

Potential impact of imposing best management practices

the relationships between related series over time. VAR models do not impose a specific structure on the system being estimated, and allow the data to ‘speak for itself’ in illustrating the patterns occurring in the industry. A VAR model is able to summarize the time-series properties and patterns occurring in past data in order to mimic the underlying data generating process as accurately as possible. Once an accurate model has been developed, shocks representing policy changes can applied to selected series and the likely impacts those shocks have on the interrelated series within the system can be observed. Therefore, the VAR economic model can provide a measure of the national impacts of these shocks on an industry level. In particular, we model the US poultry industry using a Bayesian vector autoregressive (BVAR) approach.

Data and model specification The data were obtained from US Egg and Poultry statistics for 1978–1992 (USDAERS).1 Monthly time series data for total US consumption (measured in 1000 lbs. units), production cost on a live-weight basis (cents lb 1 ), 12-city composite wholesale price (cents lb 1 ), total ready-to-cook US production (1000 lbs. units) were utilized. The selected data series summarize the aggregate dynamic characteristics of the broiler industry. The index of poultry and egg prices received by farmers was used to deflate production cost and wholesale price. The net costs of poultry and livestock manure management systems and the costs of pollution control are generally ignored or underestimated in livestock enterprise budgets. In one of the few published studies, Ashraf and Christensen (1974) suggest that manure system budgets may underestimate direct economic costs to the producer by as much as 40%. The CAST (1996) report emphasized that there are ‘no acceptable methods of calculating true costs, in which all costs associated with pollution control are factored’. The BVAR model offers useful insight into the dynamic equilibrium linkages between shocks to production costs 1 Units

are in US dollars and US or avoirdupois pounds.

linked to nutrient and manure management restrictions and industry production, wholesale prices and consumption levels. Such models can assess the impacts of policies and interventions on the collective long-term survival of these agricultural production sectors.

Preliminary tests of non-stationarity To make valid inferences regarding the behaviour of the industry over time, we must first determine if the data are stationary. Stationarity is defined as the mean and variances of the individual series and the covariance between series being independent of time. If the data are deemed stationary, then analysis with a VAR model is appropriate. If the data reject the null hypothesis of stationarity, then a vector error correction (VEC) model is indicated. The data are analyzed for stationarity using two tests, the first is a test for the existence of unit roots. In order to ensure that an autoregressive process is stationary, it is necessary that the roots of the characteristic equation are greater than 1, or lie outside the unit circle (hence the term ‘unit-root’ test). The unit root test was based on the ‘Augmented Dickey-Fuller’ (ADF) test (Dickey and Fuller, 1979). All the variables had an ADF t-test statistic greater than the critical value at the 0Ð10 level. Therefore, the null hypothesis of no unit root was not rejected indicating that all the variables are stationary in nature.2 Though the individual series exhibited stationarity, the system of interrelated series may contain cointegrating vectors and cannot be deemed stationary if cointegrating is present. Cointegration means that although many developments may cause permanent changes within a series, there is some long-term equilibrium relation tying the individual components together. If this relation exists, then it should be accounted for in constructing the time series model. Johansen’s Full Information Maximum Likelihood approach (Johansen and Juselius, 2

For a discussion of multiple-equation time-series models, testing for trends and unit roots and cointegrating and error correction models see Enders (1995).

147

148

C. S. McIntosh et al. Table 1. Testing for the number of cointegrating vectors (r) at different lag lengths (p) Lag length (p) 1 6 7

r D0

r 1

r 2

33Ð217 30Ð862 30Ð472

16Ð582 15Ð296 15Ð459

3Ð110 3Ð996 5Ð808

Critical value at 90% confidence is 49Ð925, therefore for each lag length tested, we fail to reject the null hypothesis of no cointegrating vectors (i.e. rD0).

1990) is used to derive maximum likelihood estimators of the cointegrating vectors for an autoregressive process. The null hypothesis that there are at most r cointegrating vectors in the system is tested using two likelihood ratio tests called the trace test and maximum eigenvalue test. The results of the cointegration tests are presented in Table 1. The table shows the likelihood ratio trace test results for at the most r cointegrated vectors. At all the lag lengths tested, the computed trace test statistic for, at most, no cointegrating vectors was less than the corresponding 90% critical value .tracer4,90% D 49Ð925/. Therefore, we do not reject the null hypothesis of no evidence of cointegrating vectors in the system and conclude that a VAR model is the appropriate vehicle for estimation.

Model specification Consider yt , a nð1 vector of variables which describe the structure of broiler production. The four structural variables considered are: total consumption of broiler meat (C), production cost per pound of meat produced (PC), wholesale producer price per pound (Pr), and total broiler meat produced (Pd). Assume that the dynamic behaviour of yt is governed by the following structural model: Byt DCdt CE.l/yt

1 Cut

V.u/D

.1/

where E.l/DA.L/B and yt Dn ð 1 vector of variables observed at time t B, CDfull rank n ð n matrices of coefficients A.L/Dmatrix of polynomials of order n in the lag operator L that captures

the time series properties of the broiler industry dt Dn ð 1 vector of the deterministic component corresponding to yt ut Dn ð 1 vector of structural disturbances, V.u/Dthe variance-covariance of the structural disturbances (innovations) Dn ð n covariance diagonal matrix of the structural innovations. The ut vector is also referred to as the innovation vector or vector of shocks. The vector of structural disturbances, ut , is assumed to have a mean of zero and assumed to be serially uncorrelated, mutually orthogonal and have unit variance. A reduced-form representation of the structural model shown in Equation (1) that depends only on the observable variables of yt can be obtained by pre-multiplying both sides of Equation (1) by B 1 . The autoregressive representation for the n-vector y given by: yt DB 1 Cdt CB 1 E.l/ytI CB 1 ut , yt DCŁ dt CF.l/Ł yt Cvt ,

.2/

here Cov.vt / D . The vt are mean zero, serially independent, one-step-ahead forecast errors. The reduced form in Equation (2) summarizes the sample information about the joint process of yt variables. F.l/Ł is a matrix of autoregressive parameters where Fij Ł .l/ is the i,jth element of the matrix. In the unrestricted VAR model, each variable in the system depends on lagged values of itself and lagged values of all the other variables. A common order of lag length needs to be specified for the unrestricted VAR model. We start with a maximum lag length of 12 months to capture any yearly pattern in the broiler data. The likelihood ratio test statistic (Tiao and Box, 1981) with the corresponding level of significance was estimated for all combinations of shorter versus longer lag lengths between one and 12. The results of this estimation is presented in Table 2. If the estimated c2 value had a significance level lesser than 95%, the null hypothesis was not rejected and the shorter lag length k1 was chosen. Lags longer than 12 were not considered because

Potential impact of imposing best management practices Table 2. Likelihood ratio test statistic for lag length in VAR models Longer lag .k2 /a 2 3 4 5 6 7 8 9 10 11 12 9 9 9 9

Shorter lag .k1 /a

M.k2 , k1 /b

Levels of significance

1 2 3 4 5 6 7 8 9 10 11 7 6 5 4

122Ð99 86Ð28 37Ð91 18Ð21 12Ð6 51Ð94 51Ð44 90Ð18 7Ð08 24Ð87 33Ð70 139Ð57 188Ð60 200Ð62 215Ð81

0Ð00 0Ð00 0Ð00 0Ð31 0Ð70 0Ð00 0Ð00 0Ð00 0Ð97 0Ð07 0Ð01 0Ð00 0Ð00 0Ð00 0Ð00

a The number in bold is the non-rejected lag length, for example examining the first two rows, a two period lag is not rejected when compared to a one period lag, but is rejected by a three period lag. b M.k , k / is approximately distributed as a c2 with 2 1 m2 .k2 k1 / degrees of freedom.

they would result in a large number of estimated parameters and reduce the degrees of freedom. Shorter lag lengths such as 1, and 2, were not considered as they fail to incorporate enough past information to summarize important changes in the industry. The growing cycle of broilers in the US is slightly less than 2 months, so changes in production patterns would not be reflected until the third data period. The remaining lag lengths were tested against each other to determine the appropriate lag length. Lag 10 was shown to be preferred to lag 11, but was rejected by lag 9. Lag 9 rejected all the lag lengths between 4 and 8. Therefore, the lag length was set at 9. Litterman (1984) developed a systematic method for specifying priors for a BVAR. A ‘prior’ is a set of beliefs, held by the investigator, which is imposed on the estimation. Litterman (1984) suggested using the Bayesian prior distribution on parameters of a VAR, which centers on a simple random walk process for each individual series. This is based on the assumption that the behaviour of most economic variables can be approximated as a random-walk around an unknown, deterministic drift. This prior, in its simplest form is not informative in terms of economic theory (Bessler and Kling, 1986).

The construction of a BVAR model proceeds with the specification of multivariate normal prior distribution over the coefficients Fij Ł .l/. Litterman’s prior specifies all coefficients in Fij Ł .l/ to have zero means, except for the first lag of the dependent variable which has a mean of one. The initial own lag coefficients Fii Ł , are equal to one for the series specified in logs. The F Ł ij .l/ are uncorrelated across all i and j. The random-walk prior is supplemented with additional assumptions on the form of the distribution of the prior means. Variable lags further in the past have less explanatory power than the more recent lags, thus the standard deviations decrease as the lag lengthens. The parameter l represents the overall tightness of the prior or how close all of the coefficients are to their prior mean. Low values of l imply a tight prior in which the distributions of the estimated coefficients are tightly spiked around the prior means. The decay parameter g1 determines the rate at which the standard deviations decrease on coefficients in the lag distributions. This implies that as the standard deviations become tighter around the mean on the coefficients, lags farther back in time receive less weight. The parameter g2 represents the relative tightness of standard deviations of own lags of dependent variables compared to lags of other variables in the system. Own lags of the dependent variables typically carry a weight of 1Ð0. Other variables would be assigned weights ranging from 0Ð0 to 1Ð0. A value near zero suggests that the prior mean is correct and does not allow the data to have much influence on resulting models. When other variables receive the same weight in each equation, the prior type is known as symmetric prior. The weights can be more finely tuned to each individual equation using a general prior g2 .i, j/. The parameter g2 .i, j/ reflects tightness information on coefficients of variable j in the ith equation of the VAR. Values on g2 .i, j/ range from 0 to 1. Given the parameters .l, g1 , g2 /, the Litterman prior for the standard deviation of coefficient i, j at lag 1 is given by:

dlij D

  

l lg1   lg2 sO i lg1 sO j

if iDj .4/ if i6D j

149

150

C. S. McIntosh et al.

In the prior, sO I is the standard error of the residuals from the univariate autoregressions for variable i of the lag length chosen for the VAR. The priors are scaled by a ratio of standard errors from the univariate autoregressions. The scaling is necessary so that the units in which the variables of the original series are measured do not bias the specification of the BVAR. Here, a symmetric prior was specified, where the tightness parameters for the coefficients of variable i in equation j are the same for all i and j. Next, a grid search for choosing the appropriate values of the parameters .l, g1 , g2 / was conducted based on data over the period January 1978 through December 1983. In this search the value of g ranged from 0Ð1 to 1Ð00; g1 ranged from 0Ð001 to 1Ð0; and g2 took values such as 0, 1 and 2. Out-of-sample forecasts are generated for each combination of parameter settings for the period January 1984 through December 1992. The one-step ahead forecast performance statistic as given by the Theil’s U-values was calculated for each combination of parameter settings. The combination parameter setting for .lD0Ð3, g1 D0Ð0, g2 D0Ð4/ was determined to provide the minimum Theil’s U over the out-of-sample period. This combination of parameter settings was used to specify the BVAR model. A BVAR can be further analyzed to study the structural dynamics of the system. Sims (1980) termed such an analysis ‘innovation accounting’. Innovation accounting involves obtaining the impulse response functions.

Impulse response functions Consider two variables yt and zt . Let the time path of yt be affected by the current and past realizations of the zt . The time path of zt is also affected by the current and past realizations of the yt . Let this bivariate system be represented by: yt Db10 b12 zt Cg11 yt

1 Cg12 zt 1 Ceyt

.5/

zt Db20 b21 yt Cg21 zt

1 Cg22 yt 1 Cezt

.6/

where it is assumed that both yt and zt are stationary. It is also assumed that eyt and ezt are uncorrelated white noise disturbances with standard deviations of sy and sz , respectively. Since yt and zt are allowed to affect each other, bij and gij are its coefficients that

allow the structure of the system to incorporate feedback. Equations (5) and (6) constitute a first-order VAR where the structure of the system allows yt and zt to affect each other. The fact that the yt and zt can affect each other allows us to trace out the time path of the various shocks on the variables contained in the VAR system. Plotting the impulse response functions from the system is a practical way to visually represent the behaviour of the yt and zt series in response to the various shocks. The main component of production cost is feed cost. The major determinants of feed cost are corn and soybean meal prices. Grain contributes between 60 and 70% of the total broiler ration with soybean meal accounting for between 15 and 20%. Feed costs represent as much as two-thirds of the cost of producing a live bird and these input prices are influenced by factors external to broiler production (Chavas and Johnson, 1982). Thus, production costs were assumed to be exogenous while identifying the structural model. This exogeneity makes feed cost and production costs unresponsive to shocks from other variables in the system. A production cost shock, however, has significant influence on the other variables included in this study. Rinehart (1997), noted that producers closely monitor production costs since even small cost changes make the difference between profit and loss in the broiler industry. A positive shock to cost leads to a decrease in net returns to broiler production. The imposition of nutrient management legislation would cause a significant shock to production costs (Westenberg and Leston, 1995). Under a nutrient management plan, poultry producers would be required to take greater care in the disposal of litter, much of which is currently land applied without great concern for application rate, or nutrient content of the litter. Most nutrient management plans would require an analysis of the litter for N, P and K content, an analysis of the soil upon which the litter is being applied, and an analysis of the nutrient requirements of the forage or crop being grown on the land (US Environmental Protection Agency, 1993; Bosch et al., 1997). In addition, more controlled processes of applying the manure would be required, in order to match the level of nutrients being applied to the needs of the crop.

Potential impact of imposing best management practices

Results Wholesale price and production levels were examined in response to three levels of shocks in production costs. The production cost shocks were 8, 40 and 80%. The eight percent shock corresponds to the increase in costs of dairy production found by Clouser et al. (1994), and is intended to provide an estimate of the impact of an N-based BMP. The 40 and 80% shocks are based on the assertion by Moore et al. (1995) that a P-based BMP would require 5 to 10 times more land for manure application than an N-based BMP. An increase in the amount of land required for litter application translates to increases in the cost of transportation, application and labor, so a factor of 5 to 10 times was utilized as an estimate of the impacts of a P-based BMP. Figure 1 illustrates the decline in production due to the three levels of shocks in production costs. An eight percent shock on production cost led to a 0Ð07% decline in production levels in the first month. After 12 months, the production has decreased by 0Ð2% and reaches its lowest point of 0Ð4% after 27 months. The responses to the 40 and 80% shocks were correspondingly higher, with initial decreases in production of 0Ð33 and 0Ð67%, respectively. After 27 months, production reached its lowest point with

reductions of 2Ð0 and 4Ð0%. After 48 months, fluctuations in production levels die out, with monthly variations not exceeding 0Ð02% for the 8% shock. The smoothing-out of the impulse responses by the end of 30 months indicate that the shock has only a temporary effect, and the production levels stabilize, albeit at a lower level. The shock to production costs also affects wholesale prices. During the drought of 1983–1984, production costs rose due to tight feed supplies. This situation limited production decisions which eventually resulted in high wholesale prices. Babula et al. (1990) attributed such price-increasing shocks to corn production in explaining poultry prices. The patterns of impulse responses parallel those expected where producers are pricetakers in a perfectly competitive industry. Poultry producers, having faced higher production costs, marketed birds early leading to price-depressing higher slaughter rate. More recently, increased corn price lead to an increase of farm poultry price. This is indicative of the change from an industry of many small, price taking producers to a vertically integrated industry where producers had the market power to pass on corn-based feed cost increases to consumers. Figure 2 shows a similar price response to that found by Babula et al. (1990) for the three levels of shocks in production costs. The impulse response from the eight percent

1

Percent change in production

0

–1

–2

–3

–4

–5

1

5

9

13

17

21

25

29 33 Months

37

41

45

49

Figure 1. Production level response to shocks in production cost. 8% shock ( ). 80% shock (

53

57

); 40% shock (

);

151

C. S. McIntosh et al. 4 Percentage change in wholesale prices

152

2

0

–2

–4

1

5

9

13

17

21

25

29 33 Months

37

41

45

49

Figure 2. Wholesale price response to a production cost shock. 8% shock ( ). 80% shock (

shock shows an initial increase in prices of 0Ð11% in the first month. After a slight decline in prices in the third month, prices increase in months four through 10, with the largest percentage increase (0Ð35%) occurring in the sixth month. The impacts of the larger shocks are correspondingly higher, with sixth month increases of 1Ð75 and 3Ð5% for the 40 and 80% shocks, respectively. The decline in prices in the third month may be due initial lags in the transmission of the production cost shock. After the initial decline, the price response picks up in the consecutive months with increases occurring through the 10th month. Following the 10th month, the price response is negative and eventually dies out indicating stationarity. The broiler industry conducts and monitors consumer attitudes and factors influencing chicken consumption. Consumers rank chicken’s nutritional quality along with convenience, versatility and taste as the primary factors in choosing chicken over other meats. Watts (1994), suggested that while price remains important, consumers are increasingly concerned with healthy eating and nutrition, and these factors are gaining in importance as reasons for consumers to purchase chicken. Industry analysts emphasize that quality and taste play a dominant role in food selection. Therefore, it is not inconceivable that nutrient management regulations will have a minimal impact on price levels for

53

57

); 40% shock (

);

poultry relative to competing products and can be absorbed by consumers. The behaviour of price in response to a production cost shock indicates an eventual increase in wholesale price. This suggests that the industry has the efficiency to transmit feed cost increases to consumers through increased prices. We can infer that this power over price transmission was made possible by the increased level of vertical integration that has occurred in the US broiler industry.

Policy implications and conclusions It is clear from the impulse response functions that a production cost increasing policy, such as the imposition of nutrient management legislation, would have a negative impact on poultry production levels (Figure 1). The imposition and enforcement of required BMPs would increase the overall cost of poultry production. This would, in turn, lead to higher wholesale broiler prices. These wholesale price impacts, however, appear to be transitory in nature. In the case of both the production level and wholesale price response, the impacts from the shock to production costs were quite small on a percentage basis but proportionately larger for P-based BMPs. This indicates that the

Potential impact of imposing best management practices

impact of a cost-increasing event, such as the imposition of nutrient management legislation, would not cause major changes in poultry production, or wholesale prices. Rinehart (1997) has expressed concern that US producers cannot afford to stand by and be put at a disadvantage in the world market due to excessive regulation. The integrated broiler industry model confirms that the impact of nutrient management regulations are not a significant barrier to expansion and continued survival of the industry. The price trends and rate of adjustment in broiler prices in response to external cost shocks identified in the BVAR model have useful implications for manure management regulatory policy. The CAST (1996) report on integrated waste management noted the importance of increased efforts to educate poultry and livestock producers due to intensified regulatory pressure and external public concerns. The dynamic model confirms that both producers and consumers can adjust to the long term impacts of shocks to production costs associated with nutrient and manure management strategies. These results are consistent with a voluntary approach to implementing best management practices when livestock producers can gradually adopt and adjust to standards developed by industry experts. The National Pork Producers Council has implemented an Environmental Assurance Program to demonstrate the payoffs from voluntary adoption of nutrient management plans for integrated pork operations. The results presented here indicate minimal long-term industry impacts and may encourage producers to implement voluntary approaches.

References Ashraf, M. and Christensen, R. L. (1974). Analysis of impact of manure disposal regulations on dairy farms. American Journal of Agricultural Economics 56, 331–336. Babula, R. A., Bessler, D. A. and Schluter, G. E. (1990). Poultry-related price transmissions and structural change since the 1950s. Journal of Agricultural Econonomics Research 42, 13–21. Bailey, G. W. and Waddell, T. E. (1979). Best management practices for agriculture and silviculture an integrated overview. In Best Management Practices for Agriculture and Silviculture (R. C. Loehr, D. A. Haith, M. F. Walter, and

C. S. Martin, eds), pp. 35–56. Ann Arbor: Ann Arbor Science Publishers Inc. Bessler, D. A. and Kling, J. L. (1986). Forecasting vector autoregressions with Bayesian priors. American Journal of Agricultural Economics 68, 144–51. Bosch, D. J. and Pease, J. W. (1991). Economic Impacts of Manure Application Restrictions on Dairy Farms. Department of Agricultural and Applied Economics, Virginia Tech. Bosch, D. J., Zhu, M. and Kornegay, E. T. (1997). Economic returns from reducing poultry litter phosphorus with microbial phytase. Journal of Agricultural and Applied Economics 29, 255–266. Carson, J. B. and Smeltz, W. (1993). Pennsylvania Farmers’ Association Welcomes Passage of Nutrient Management Legislation. PR Newswire Association, May 5. Chavas, J. P. and Johnson, S. R. (1982). Supply dynamics: the case of US broilers and turkeys. American Journal of Agricultural Economics 64, 558–564. Clouser, R. L., Mulkey, D., Boggess, B. and Holt, J. (1994). The economic impact of regulatory decisions in the dairy industry: a case-study in Okeechobee County, Florida. Journal of Dairy Science 77, 325–332. Council for Agricultural Science and Technology (CAST) (1996). Integrated Animal Waste Management. Task Force Report No. 128, November. Ames, IA, CAST. Dickey, D. A. and Fuller, W. (1979). Distribution of the estimates for autoregressive time series with a unit root. Journal of the American Statistical Association 74, 427–431. Dickey, D. A. and Fuller, W. A. (1981). Likelihood ratio statistics for autoregressive time series with a unit root. Econometrica 49, 1057–1072. Edwards, D. R., Daniel, T. C., Scott, H. D., Murdoch, J. F., Habiger, M. J. and Burks, H. M. (1996). Stream quality impacts of best management practices in a Northwest Arkansas basin. Water Resources Bulletin 23, 499–509. Enders, W. (1995). Applied Econometric Time Series. New York: John Wiley and Sons. Gerstenzang, J. (1997). Potomac River is paying the price for nations chicken appetite: proliferation of poultry farms along waterway’s banks is reversing a comeback from decades of pollution. The Los Angeles Times: Home edn, Thursday, April 17, 1997. Johansen, S. (1988). The mathematical structure of error correction models. Contemporary Mathematics 80, 359–386. Johansen, S. and Juselius, K. (1990). Maximum likelihood estimation and inference on cointegrating with application to the demand for money. Oxford Bulletin of Economics and Statistics 52, 169–210. Leavenworth, S. (1996). Farmers fight neuse plan. The News and Observer, Final Edn, Raleigh, NC, August 30, 1996. Litterman, R. B. (1984). Forecasting and policy analysis with Bayesian vector autoregression

153

154

C. S. McIntosh et al. models. Federal Reserve Bank of Minnesota Quarterly Review Fall, 30–41. Madison, R. J. and Brunett, J. O. (1985). Over View of the Occurrence of Nitrate in Groundwater in the United States. US Geological Survey: Water Supply Paper 2275. Moore, P. A., Jr, Daniel, T. C., Sharpley, A. N. and Wood, C. W. (1995). Poultry manure management: environmetnally sound options. Journal of Soil and Water Conservation 50, 321–327. Patterson, P. H. (1996). Nutrient management: poultry manure utilization. Zootecnica International March, 46–47. Rinehart, K. E. (1997). Integrator’s perspective. Broiler Industry 60, 30–34. Sims, C. A. (1980). Macroeconomics and reality. Econometrica 48, 1–48. Soil Conservation Service (SCS) (1987a). Waste Utilization. Section IV, Field Office Technical Guide, US Department of Agriculture, Little Rock, AR. Soil Conservation Service (SCS) (1987b). Pasture and Hayland Management. Section IV, Field Office Technical Guide, US Department of Agriculture, Little Rock, AR.

Soil Conservation Service (SCS) (1992). Nutrient Management. Section IV, Field Office Technical Guide, US Department of Agriculture, Little Rock, AR. Tiao, G. C. and Box, G. E. P. (1981). Modeling multiple time series: with applications. Journal of the American Statistical Association 76, 802–816. US Department of Agriculture, Economic Research Service (1960–1992). US Egg and Poultry Statistics. US Environmental Protection Agency (1993). Guidance Specifying Management Measures for Sources of Nonpoint Pollution in Coastal Waters. Publication Number 840-B-92-002, Office of Water. Washington DC: US. Government Printing Office, January 1993. Warrick, J. and Goodman, P. S. (1998). EPA plans to regulate livestock waste: clean water act would be applied. Washington Post March, A01. Watts, G. B. (1994). Broiler markets and strategies for the 2000s. Broiler Industry 57, 26–40. Westenbarger, D. A. and Letson, D. (1995). Livestock and poultry waste-control costs. Choices 10, 27–30.