An Analytical Review of a Voluntary Herd Retirement and Export Subsidy Plan for Dairy Producers

Transcription

1 An Analytical Review of a Voluntary Herd Retirement and Export Subsidy Plan for Dairy Producers Charles Nicholson, Ph.D. & Mark Stephenson, Ph.D. 1 August, 2007 One of our colleagues 2 has characterized the milk price problem as having three aspects: price volatility, price predictability, and price adequacy. Volatility really has to do with the dramatic price swings that have become part of the dairy industry during the last 15 years (illustrated by the marked variation in the Class III price in Figure 1). High milk price years, like 2004 and 2007, are great recovery years if you can make it through the low price troughs, like 2002 and Price volatility might not be so bad if it were predictable that is, if dairy producers and processors knew with a high level of certainty what prices were going to be well in advance. Then, business plans could be altered to accommodate the price volatility. Finally, price adequacy is an issue. If the average price is too low to cover long-run average costs, then a producer will be out of business regardless of the peaks and troughs Figure 1. Manufacturing Grade Milk at 3.5% fat test Class III Support Price $/cwt, 3.5% fat The authors are Senior Research Associate and Senior Extension Associate respectively with the Cornell Program on Dairy Markets and Policy at Cornell University. 2 Andrew Novakovic, Director of the Cornell Program on Dairy Markets & Policy

2 Many dairy producers have voiced their concern about the recent milk price volatility and expressed a desire to find policies and programs to reduce it. Earlier, we analyzed 3 a Refundable Assessment plan for dairy producers at the request of the Milk Producers Council of California. Our analysis, using a dynamic national model developed at Cornell, indicates that milk price volatility can be reduced by use of national dairy policy although more in-depth analyses would be appropriate to address the details of implementing a program and determine its likely regional impacts. We found that reduction in price variation could be most effectively achieved by levying a one dollar per hundredweight of milk assessment against all milk producers and dividing the entire pool of money collect among farms that did not increase production more than an annual growth rate of three percent. The Refundable Assessment plan has the desirable attributes of transparency, simplicity and potency in moderating milk price volatility. It is a program that can be set in place and then effectively operates on auto-pilot. However, a one dollar assessment even if refundable has seemed like a lot of money to dairy producers. Also, the program penalizes rapid growth an attribute that some regions of the country desire to have happen. Given the size of the assessment, some analysts have also suggested that farms will manipulate ownership arrangements to avoid not receiving the payment on at least some of their milk. As a result of these perceived limitations of a refundable assessment program, Dairy Farmers Working Together (DFWT), a group with roots in Vermont, requested that we look at an expanded version of a Cooperatives Working Together (CWT) type of program with dual goals of moderating price volatility and price adequacy. Cooperatives Working Together The CWT program was begun in June of It was conceived by the National Milk Producers Federation (NMPF) as a voluntary program to increase demand by subsidizing exports of dairy products, and reduce supply by producers submitting bids to decrease milk production or whole-herd buyouts. It was initially estimated that the program would require a 17.9-cent assessment per cwt of milk to effectively fund the goals of the program. Because it is a voluntary program, buy-in at this level was too low and the program began with a 5-cent assessment paid by about 68 percent of the nation s milk. The decreased production option was dropped in favor of concentrating on herd retirement and export subsidy in the second year of the program. In July of 2006, the CWT program increased the assessment to 10 cents per cwt to fund additional export subsidies and herd retirements. The CWT program has set export price benchmarks for cheese and butter, based on the Chicago Mercantile Exchange (CME), at $1.30 per pound. Should the CME price fall below this level, CWT will begin actively soliciting bids for exports of these products although they will accept bids and may award bonuses at any price level. 3 An Analytical Review of a Refundable Assessment Plan for Dairy Producers. April,

3 Four rounds of herd retirement bids have been conducted since the program began. In the first year, just short of 33,000 cows went to slaughter under the program. Round 3 removed the most cows with slightly more than 64,000 cows. The average bid accepted in Round 3 was $6.75 per cwt of milk production and in the most recent round, 53,000 cows were slaughtered at an average bid of $5.50 per cwt of milk production. NMPF has estimated the impact of their CWT program as high as 60 cents per cwt on a 5 cent investment. Analysis by Scott Brown 4 of FAPRI has estimated the return at 40 cents per cwt for the 5 cent investment. The actual dollar impact of CWT may be debatable but it has been an estimable program if for no other reason than the cooperative effort that has been demonstrated by America s milk producers to deal with price level and volatility. The CWT program may be best described as a cartel where firms collude to reduce output and increase prices. What might seem to be anti-competitive behavior is allowed for cooperatives under the Capper-Volstead Act. Cartels can be a very effective price enhancement tool (although illegal without explicit exemptions in most countries), but they can suffer from a free-rider problem. It is unlikely that there are many investments on a dairy farm where a 5-cent expenditure will return cents in payback. When there are many firms involved in a cartel, a single producer understands that if everyone else contributes their nickel (or dime) and they don t, then the benefits of higher milk prices accrue to everyone and the free-rider is even better off with the higher prices and their nickel (or dime). Because of this incentive to opt out of a voluntary program, most cartels of this nature will self-destruct over time. A Mandatory CWT Program DFWT asked us to analyze the impacts of a CWT-type program that was altered somewhat from the current program. For one thing, the program would no longer be a voluntary producer effort. The modified program would result from national policy and because the benefits would accrue to all traditional dairy producers, non-organic producers would contribute to the funding of the program. Also, DFWT could replace the current Milk Income Loss Contracts (MILC) and the Dairy Export Incentive Program (DEIP). They estimated that the budget savings to the federal government from not operating these programs would amount to about $250 million per year. We were asked to model a program with access to $250 million per year from taxpayer revenues and various levels of producer revenues from a mandatory checkoff program. The dual goals of such a program would be price stabilization and price enhancement. We modified the Cornell Dynamic National Dairy Sector Model to assess the potential for success of such program goals given the constraints of program funds

4 The Cornell Dynamic National Dairy Sector Model (CDNDSM) This CDNDSM is one of a suite of dairy sector models developed at Cornell since the mid-1990s (see Appendix Table). This model builds upon a previous conceptual commodity model (Sterman, 2000) and the dairy industry price determination model in Nicholson and Fiddaman (2003). The model has been previously used to assess the market impacts of promoting high-protein whey products (Nicholson and Stephenson, 2006) and impacts of generic advertising expenditures for fluid milk and cheese (Nicholson and Kaiser, 2007). The model is a dynamic monthly national model that simulates dairy market outcomes from 2004 to It is formulated as a system dynamics model, meaning that it explicitly represents stock-flow-feedback processes 6, and focuses on the endogenous processes that generate price variation rather than on exogenous shocks (like weather-induced supply reductions or changes in EU policy). It includes a total of 17 final and intermediate dairy products (intermediate products are those dairy products used in the manufacture of other dairy products, such as nonfat dry milk (NDM) used in cheese manufacturing). Final products are those used by non-dairy manufacturers (e.g., other food processors) or final consumers. Perishable products such as fluid milk, yogurt and ice cream are flow variables for which production is equal to sales. Commercial inventories of storable commodities such as butter, cheese, dried milk and dried whey are stock variables, where production increases inventories and sales reduce them. Increases in commercial inventories of these products result in decreases in prices. The model includes the use of skim milk and cream components to capture component balance. On the production side, supply response occurs through changes in both productivity (milk per cow) and changes in the number of cows. Productivity responds with a delay to changes in the milk price in the short run: complete response to a step change in the milk price relative to a reference price occurs within about 3 months. The model represents the number of cows and replacement animals (heifers) as well as sales of bull calves. The number of cows is determined by the reproduction rate (based on biological parameters and a calving interval) and the culling rate. The culling rate responds to a long-run expected net margin (milk price less production costs), where the expectation of future net margin is a function of past net margin values. The culling rate is partitioned into involuntary culls and voluntary culls. Eleven of the 17 dairy products are considered tradable, and US imports and exports of these products are explicitly included. For the purposes of this project, the most important tradable products are American cheese, NDM and butter. The model does not represent complete world trade flows of dairy products. Rather, it takes 2004 export values and world prices as a baseline, and modifies US imports and exports on the basis of changes induced by developments in US dairy product markets. For example, an in- 5 Although the model is simulated monthly, it is based on annual production and consumption data and ignores seasonal patterns in demand and milk production. 6 Mathematically, the model is a system of nonlinear differential equations solved by numerical integration

5 crease in US exports of cheese (for a given level of export demand) would result in a decrease in the world cheese price, which may then also increase US imports of cheese. Parameter values for behavioral responses to prices and inventory levels and the initial stock values are from previous dairy industry models (Nicholson and Bishop, 2004; Nicholson and Fiddaman, 2003), data from the Agricultural Marketing Service of the US Department of Agriculture, and an extensive network of industry contacts (Cornell Program on Dairy Markets and Policy, 2006). The model base year is Underlying growth in demand for dairy products (i.e., outward shifts in the demand curves) in response to increases in household income and population are exogenous are from Nicholson and Stephenson (2006) and Schmit and Kaiser (2006). The data used for the model do not account for changes in world market conditions and US exports since 2004, nor for specific milk supply shocks. Policy-Related Model Elements Because this is a national model, all milk is assumed to be regulated under the classified pricing system operated by Federal Milk Marketing Orders, and the model explicitly includes the pricing formulas and timing of price announcements. The Dairy Price Support Program is explicitly represented through the willingness of the CCC to purchase American cheese, butter or NDM products when prices fall below purchase prices. Dairy trade policies include detailed product-specific representation of import quotas (both within- and over-quota imports as represented by TRQs) and export subsidies for American cheese, butter and NDM are included with both quantity and expenditure limits. Payments to dairy farmers under the MILC program are included, with payments per cwt equal to the difference between the MILC target price of $16.94/cwt less the actual Class I mover plus the Class I differential at Boston ($3.25/cwt) when the actual price is less than the target times the payment proportion (0.34) times the proportion of milk produced below the cap of 2.4 million lbs (assumed to be a constant value of 0.67). The model simulations do not include the effects of previous rounds of the voluntary CWT program. Modifications to Represent the DFWT Proposed Program The key elements of program design for the DFWT proposed program and how the CDNDSM has been modified to address them are as follows: How much money will be collected, how, and from whom? The model allows simulation of a wide variety of combinations of a mandatory assessment per cwt on all non-organic milk and a lump-sum monthly contribution from the government. For scenarios in which government funds are provided, we assume that the government provides money for the program regardless of whether it is currently active. The values of these parameters can be varied to assess with combinations work more effectively to increase and(or) stabilize milk prices. The funds collected are assumed to available to spend over a time horizon determined to be appropriate (that is, funds collected in a given year don t have to be - 5 -

6 spent in that same year, although this may not be a good assumption for government funds). How will the money collected be spent? This is one of the more challenging aspects of the analysis due to the large number of ways in which the funds could be expended. The model representation must simplify from a number of decision factors that might be used in the real world, and uses a logical, hierarchical set of decision rules. First, the currently available funds (the sum of all monies collected less all monies spent to date) are calculated. Under certain conditions, the money will be spent (more on this below). When those conditions are met, the model assumes that the available money would be spent over the course of the following 6 months (this assumption can easily be changed). This monthly amount is called the Program Funds to be Spent. This money will then be allocated to either herd retirements or export subsidies. We explored two possible rules for when monies will be spent. The first rule is to establish what we call a Program Activation Price (PAP). When the current all-milk price is below this PAP, the program will become active and begin spending money on herd retirements and export subsidies. The extent to which money will be spent depends on a non-linear function of how close the current price is to the PAP. That is, if the current all-milk price is close to the PAP, the program will spend money less aggressively to avoid overshooting in the marketplace. When the current price is well below the PAP, the program will fully expend money available to it, consistent with the timing discussed above. The PAP should not be confused with a price target. It is simply a price that is used to determine when the program will be active and how aggressively money will be spent. The second rule, which has been explored only in a limited fashion is that the program becomes active when the expected trend in milk prices is negative. An initial version of this rule was examined and resulted in increased price variability and ever-increasing prices, so was dropped from subsequent analyses. How does the model allocate money to export subsidies? The model assumes a maximum allowable proportion of available funds can be spent on export subsidies and allows expenditures up to this amount (but expenditures can be less than this amount). A fixed proportion of the total funds for export subsidies is assumed to be available for each of the three products (American cheese, NDM and butter), independent of current world market conditions. Actual expenditures on export subsidies for these three products depend on whether the current US price is above a specified target (if the US price is above the target, no money is spent) and whether an export subsidy is required (if the US price is equal to or less than the world price, a subsidy is not required). If the US price is below the specified target price and the US price is above the world market price, the money available for export subsidies on each product is divided by the unit subsidy required (i.e., the difference between US and world price) to determine the quantity of subsidized exports

7 How does the model allocate money to herd retirements? The money allocated to herd retirements is simply the amount of Program Funds to be Spent less the sum of actual expenditures on export subsidies on the three products. Thus, in each period, the Program Funds to be Spent will be expended. Herd retirements are modeled rather simply; we do not model individual farm participation decisions. Rather, we assume that the total number of cows removed in a given month depends on the money available for herd retirements. The amount of cwt of milk removed each month is the total amount of funds to be expended divided by an average Bid Price per cwt for which producers are willing to liquidate animals. The Bid Price per cwt increases as the all-milk price increases, with the initial value based on the actual average bids from the 2007 CWT program. The number of cows to be removed is calculated as the number of cwt of milk to be removed divided by the average milk per cow. (Note that this may overestimate the reduction in milk supplies if participants have lower than average cow productivity.) It is well known that herd retirement programs end up paying some proportion of their funds to herds (cows) that would have exited the industry anyway. To address this, we reduce the effective number of culled cows by what Pagel (2005) estimated to be the rate of involuntary culls, 15% of animals at an annual rate. How are cull cow prices represented? A concern of other livestock producers is the impact that a herd retirement program has on livestock prices. We include a simplified representation of this effect to help assess the likely changes in livestock prices under the program. In simple terms, cull cow and calf prices are responsive to calf production and cow culling rates. When culling rates are greater than a baseline culling rate, cull cow prices fall and when culling rates are lower than the base value, cull cow prices increase. A similar effect is modeled for calf prices. We have ignored growth in beef demand, so our simulated beef prices are likely to be lower than those that would actually occur in response to the program. Scenarios Analyzed and Outcomes of Interest The model structure is flexible enough to allow a wide variety of scenarios to be analyzed quickly, to conduct what are called sensitivity analyses (that determine how important changes in specific assumed parameter values are to the simulated outcomes) and to identify best case combinations of program operational parameters (such as the amount of the assessment or government expenditures). We have used all of these approaches to generate the results described below. It is quite common in dynamic simulation models to simulate a baseline scenario that represents what would happen under a given set of assumptions if the current programs are maintained. Alternative scenarios that indicate what happens if a program like DFWT is implemented are then simulated, and the impacts of the program are the differences between the simulations with the program and the baseline scenario. That is, the differences between the simulations are more important than the actual levels - 7 -

8 of variables simulated. Thus, we report mostly the differences between the baseline and the DFWT scenarios under different assumptions. A summary of four scenarios that account for different assumptions about the level of assessments and government expenditures is provided in Table 1. Baseline Scenario Table 1. Summary of Scenarios Analyzed MILC Active under current program operating procedures Assumptions DEIP Active under current program operating procedures $/cwt None Assessment Government contributions $ million / year None No assessment, government contribution Not active Not active cent assessment, no government contribution Not active Not active cent assessment, no government contribution Not active Not active "Best case" with government contributions Not active Not active "Best case" without government contributions Not active Not active * Identified through sensitivity and other analyses as the scenario resulting in the lowest price variation after an implementation period and the highest producer price per cwt. It is also of great importance to understand that all models are designed for a specific purpose. The purpose of this model is to provide an initial assessment of the ability of the proposed DFWT program to enhance and(or) stabilize milk prices. In particular, the CDNDSM is not a price forecasting model. The prices generated in the baseline scenario reflect the evolution of prices in the US dairy industry over time given 2004 market conditions and the evolution of endogenous decision making processes in the industry (i.e., supply and demand responses to price signals). The baseline scenario does not account for other exogenous impacts on US dairy markets (supply shocks, export demand shocks, world price changes) during 2004 to 2007 that would be important for accurate forecasts of future prices. Thus, it is important to interpret the comparisons of the baseline with the DFWT scenarios as an initial assessment of the ability of the program to enhance prices and(or) stabilize them in the presence of endogenous price and pro

9 duction cycles, not as an assessment of the program under all possible US and world market conditions. It is also important to clearly state what outcomes will be examined to determine success of the program. This is important in part because many policies and program will result in a mixed record with regard to achieving a set of objectives. (For example, it might be possible to increase average all-milk price, but only with an increase in price variation.) It is also relevant to distinguish between program inputs (for example, expenditures) and program outcomes (like the all-milk price). The table below summarizes the inputs and outcomes that are the focus for the discussion of our results. It is important to note that the outcomes of interest should be evaluated from the time of program implementation and, as noted earlier, as comparisons to the Baseline. For evaluation of the impacts on price stability, it is relevant to allow a period of time for adjustment after program implementation before examining how the program influences price variability. This is appropriate because the proposed program typically results in initial increases in all-milk prices (a desirable outcome for dairy producers) but these increases would also be counted as price variability. Based on the response of the allmilk price to the program, we somewhat arbitrarily chose 24 months as the adjustment period after which price variation would be evaluated. Table 2. Program Inputs and Outcomes for Analysis of Proposed DFWT Program Program Inputs Program Outcomes Difference in average all-milk price from Baseline Expenditures on herd retirements scenario after program implementation Average absolute deviation from average all-milk Expenditures on export subsidies price less assessments after program implementation and an adjustment period Difference in producer net revenues (milk and livestock revenues less production costs) from Baseline Total program funds available after program implementation Differences in total cows culled from Baseline after program implementation Differences in cull cow prices from Baseline after program implementation Differences in net imports from Baseline for subsidized products after program implementation Differences in Class III and IV prices from Baseline after program implementation - 9 -

10 Results Hundreds of scenarios were run with the model to determine the impacts of many factors and assumptions, but we examined most closely the level of the PAP, the maximum allowable proportion to be devoted to export subsidies and the source of the funds. One of the scenarios made about 20 million taxpayer dollars available every month as the only source of funds. This represents the $250 million annually estimated by DFWT that would be saved from the federal budget by dropping the MILC and DEIP programs. Two other scenarios were to assess dairy producers 10 or 15 cents per cwt to fund the program and assume that there were no government funds available. Finally, sensitivity analysis was performed using the model to select two variability-minimizing cases (which have been designated Best Cases ). One of these scenarios allowed a combination of government funds (up to $20 million per month) and producer checkoff funding up to 15 cents per cwt. The other best case scenario did not allow any government funds. The best case scenario was determined as the least volatile combination of government and private dollars but also provided a higher average all-milk price. With 200 scenarios run for each of the best case programs (with and without government support), Figures 2 & 3 show how the best case scenario was selected. On the vertical axes is a measure of volatility while on the horizontal axes is a measure of the difference in the average all-milk price after program implementation compared to the initial all-milk price. Figure 2. Increase in Average All-Milk Price During Program Operation and Average Absolute Deviation for N=200 Model Simulations with Government Contributions Average Absolute Deviation, $/cwt Point selected for best case Increase in Average Price Over Initial Price (PAP increment), $/cwt

11 Figure 3. Increase in Average All-Milk Price During Program Operation and Average Absolute Deviation for N=200 Model Simulations without Government Contributions Average Absolute Deviation, $/cwt Point selected for best case Increase in Average Price Over Initial Price (PAP increment), $/cwt Figure 4. Producer Price Difference From Initial Value with DFWT Program, Various Assumptions 1.60 Program Start Adjustment Period Variability Evaluation Period $/cwt Month Baseline Best Case with Govt Best Case no Govt No Assessment, Govt=20 mil No Govt, Assessment = 10 No Govt, Assessment =

12 Figure 4 shows the difference in producer prices from the initial 2004 year prices. Appendix Table 2 displays specific values of the various scenarios for the for the five years of the program simulation and the three years after the adjustment period. You will note that the first 12 months show no price difference in those 12 months as they are identical to the initial year. After the first year, all scenarios begin to deviate first lower than 2004 and then hitting a peak about $1.40 higher into We assume that the program scenarios all start at the beginning of For purposes of measuring price volatility, we assume that the program has a phase in period (shown in Figure 4 as the Adjustment Period) when they are first being implemented. Price volatility measures are calculated during the final three years after the Adjustment Period. Producer prices are calculated as a national federal order blend price plus over-order premiums minus producer assessments, but also include MILC payments in the Baseline scenario. The dark blue line in Figure 4 shows what the model is projecting if policies stay the same. We would expect to see continued price volatility with an oscillation period of just less than three years between peaks or troughs. This is the do nothing scenario referred to in this and all subsequent figures as the Baseline. The price pattern that arises through implementation of the program using only the $20 million per month funding that might be available from government budget savings is shown by the yellow line. This scenario generates the highest average price (38 cents per cwt higher than the Baseline) for producers with somewhat less volatility than the Baseline. If only a mandatory 10 or 15 cent per cwt producer assessment were used to fund the program, somewhat lower volatility and somewhat higher average prices than the Baseline are predicted to result (light blue and green lines in Figure 4). But the use of only producer assessment funds results in a more volatile and a lower average price than the government-only funded program. In fact, the average all milk price over the program lifetime under this scenario is about 18 cents less per cwt than the yellow line. This perhaps isn t a surprising outcome as the 10 cent assessment generates about $70 million less in funds available to operate the program over the three years in the Variability Evaluation Period. As previously stated, the best case scenarios are selected from hundreds of model simulations using various combinations of taxpayer and producer funds. Figure 5 below shows N=200 combinations of taxpayer and producer funds and the solid colors indicate where 50%, 75%, 95% and 100% of the time the outcomes lay

13 Figure 5. Producer Price Difference from Initial Model Price for N=200 Model Simulations Figure 6. Average Absolute Deviation from Producer Price After Program Implementation and a 24-Month Adjustment Period for N=200 Model Simulations $/cwt $/cwt

14 The simulation runs also allowed us to calculate a measure of volatility as an average absolute deviation from producer prices after the adjustment period of 24 months. It is this measure that was used to choose the best case combination and level of federal and producer dollars to fund the program to achieve minimum price volatility. The red line in Figure 6 shows the volatility in the best case scenario while the black line shows the baseline volatility. The no government funding and the producer assessment scenarios are displayed on Figure 6 as a grey and black line respectively. This figure indicates that most scenarios, operating the DFWT program under the rules outlined on pages 5-7 above, will result in less volatility than no program. The best case scenario would markedly reduce price variation during the last three years of program operation. However, it is important to note that there are some combinations of producer assessments, government funding, program activation price and maximum percentage of program funds allowed for export subsidies under which the program would cause even more volatility. Although based on the sensitivity analyses shown above (Figures 5 and 6) there appears to be a relatively small probability of a bad outcome, DFWT is not a fail-safe program to achieve the objective of reduced price fluctuations. The no government funding with producer assessment scenarios are displayed on Figure 6 as a grey and black lines. Presumably, dairy farmers care about more than just volatility and milk price levels. Differences in producer net revenues from the baseline can also be calculated and give a more complete understanding of the changes in farm profitability than milk price levels alone. Net producer revenues take into account whether farms are producing more or less milk (as well as the milk price), sources of other income (like cull cows, etc.) and other costs (such as variable costs of production). Figure 7 shows the changes from the baseline for the three scenarios. Figure 7. Producer Net Revenues Difference from Baseline with DFWT Program, Various Assumptions 120,000,000 Program Start Adjustment Period Variability Evaluation Period 100,000,000 80,000,000 60,000,000 $/month 40,000,000 20,000, ,000,000-40,000,000 Month Baseline Best Case with Govt Best Case without Govt No Assessment, Govt=20 mil No Govt, Assessment = 10 No Govt, Assessment =

15 On average, all of the scenarios provide higher net revenues. The Best Case averages an additional $32-33 million per month, the government funded scenario contributes $35 million per month, and the producer assessment scenario provides an additional $27 and $39 million per month for the 10 and 15 cent assessment respectively to U.S. dairy farms. There are two uses of the CWT program funds: herd retirement and export subsidy. Figure 8 shows how the funds are expended under the best case (lowest volatility) scenarios. During the adjustment period, the program struggles under the rules (page 6) to subsidize exports at varying levels and in particular, on expenditures for herd retirement. After the adjustment period, export subsidies begin to settle out at about $2 million per month with government support and $3 million without. Herd retirement expenditures, while still variable, begin to average something like $15 million per month in both cases. These levels of expenditures on the two branches of the program provide the least variable farm milk prices. Figure 8. Expenditures on Export Subsidies and Herd Retirement Under Best Case Scenarios, With and Without Government Funding 25,000,000 Program Start Adjustment Period Variability Evaluation Period 20,000,000 $/month 15,000,000 10,000,000 Expenditures on Herd Retirement with Govt Expenditures on Export Subsidies with Govt Expenditures on Herd Retirement No Govt Expenditures on Export Subsidies No Govt 5,000, Month Farm milk prices (Figure 4) are more stable with the best cases program implementation. For manufacturers, class prices also need to be considered. Figures 9 and 10 show what is happening with the Class III and Class IV milk prices over the timeline. Please remember that these figures represent differences from the baseline, which was the most volatile case shown. There is a sharp, initial drop in Class IV price after the program begins that is indicative of the loss of the DEIP. The Class III price peaks

16 showing as a difference from baseline in Figure 9 around month 60 and 96, are really helping to smooth the milk price from its inherent volatility. Figures 9 and 10 indicate that manufacturing milk prices will not necessarily move in lockstep. At times when the Class III milk price is the biggest positive deviation from the baseline, around month 96, the Class IV milk prices are almost lower than baseline levels. Some of these differences are triggered by the product being exported more or less heavily in the DFWT program. Figure 9. Class III Price Difference with DFWT Program, Various Assumptions 1.00 Program Start Adjustment Period Variability Evaluation Period $/cwt Month Baseline Best Case with Govt Best Case, no Govt No Assessment, Govt=20 mil No Govt, Assessment = 10 No Govt, Assessment =

17 Figure 10. Class IV Price Difference with DFWT Program, Various Assumptions 0.50 Program Start Adjustment Period Variability Evaluation Period 0.25 $/cwt Month Baseline Best Case with Govt Best Case No Govt No Assessment, Govt=20 mil No Govt, Assessment = 10 No Govt, Assessment = 15 Dairy policy can bring about unexpected friends or foes. When the dairy termination program was slaughtering cows in the 1980s, beef producers were notably upset at the decrease in their beef price. With a dairy program that is actively encouraging herd retirement and cow slaughter, it is easy to imagine that livestock producers would be skeptical of this program as well. Figure 11 shows that the DFWT program culls somewhere between 15,000 and 25,000 cows per month

18 Figure 11. Cull Cows with DFWT Program, Various Assumptions 35,000 Program Start Adjustment Period Variability Evaluation Period 30,000 25,000 Animals/Month 20,000 15,000 10,000 5, Month Baseline Best Case with Govt Best Case no Govt No Assessment, Govt=20 mil No Govt, Assessment = 10 No Govt, Assessment = 15 Figures 12 and 13 demonstrate that the cows taken out under the CWT program may not actually be a problem for livestock producers. In fact, because the CWT program, through herd retirement and export subsidies, does raise the average price to producers, they desire to produce more milk. The only way to do that is by reducing voluntary culling in herds not participating in the retirement program. In particular, the best case scenarios cull fewer total cows (herd retirement, voluntary and involuntary culling) in Figure 12 and provides a higher average cull cow price than the baseline and cull cow price variation due to the program is less than 2 percent of the assumed initial cull cow price

19 Figure 12. Total Cull Cows with DFWT Program, Various Assumptions 530,000 Program Start Adjustment Period Variability Evaluation Period 520, , ,000 Animals/Month 490, , , , , , , Month Baseline Best Case with Govt Best Case no Govt No Assessment, Govt=20 mil No Govt, Assessment = 10 No Govt, Assessment = 15 Figure 13. Cull Cow Price Difference from Baseline with DFWT Program, Various Assumptions 20 Program Start Adjustment Period Variability Evaluation Period $/animal Month Baseline Best Case with Govt Best Case no Govt No Assessment, Govt=20 mil No Govt, Assessment = 10 No Govt, Assessment

20 All program scenarios provide incentives for somewhat less milk production than the Baseline. Because of this, there are other unexpected results from operating this kind of CWT program. Domestic prices for dairy products rise giving rise to higher farm prices the U.S. markets look like a better destination for other country s products. The model explicitly accounts for the caps on U.S. imports and tariff rate quotas negotiated under the previous round of the World Trade Organization (WTO). Although we are exporting dairy products using a subsidy program, we have also given up the DEIP and are importing more dairy products than we do with no CWT program in the baseline. Figure 14 shows that net imports (imports minus all exports) actually increase after program implementation. Figure 14. Annual Change in Net Imports from the Baseline After Program Activation. 70,000 Annual cwt of Net Imports, Various Products 60,000 50,000 40,000 30,000 20,000 10,000 0 American Cheese Other Cheese Whey Products Milk Powders Butter

21 Summary Currently, Cooperatives Working Together (CWT) is operating a voluntary program to collect an assessment from dairy producers to subsidize exports of dairy products and to occasionally accept bids from dairy producers to retire their herds from production. This program may suffer from the free-rider problem going forward and it may not be able to operate at the optimal level from lack of adequate funding. Dairy Farmers Working Together (DFWT) proposes to implement a mandatory CWT-like program, possibly using both taxpayer funding from the federal budget that would be saved by the cessation of the Milk Income Loss Contracts (MILC) and the Dairy Export Incentive Program (DEIP) and (non-refundable) assessments on dairy producers. We have modeled this program using the Cornell Dynamic National Dairy Sector Model (CDNDSM). This is not a price projection model. It does an excellent job of using underlying biological, institutional and behavioral factors to project the innate volatility in milk prices. To capture all of the highs and lows, we would also need to be able to project shocks to the economic system which, by definition of a shock, is not possible to do. Therefore, we have not modeled recent known shocks like the impacts of new overseas demand on milk prices. All scenarios are compared to a baseline outcome. Our initial assessment shows that such a program can reduce milk price volatility and increase average milk prices to dairy producers. More importantly, it can increase producer net revenues while reducing price volatility. Previous work looking at a refundable assessment program showed that price volatility can also be greatly diminished by collecting a sizable assessment from producers ($1.00 per cwt) that would be paid back if farm milk production levels did not exceed an annual 3 percent growth rate. The DFWT proposal is a more complicated program to administer than the refundable assessment program. If not properly operated, the expanded CWT program could actually yield increased volatility. In particular, we noticed more volatile prices using a simple rule of activating the export subsidies and herd retirement when there was a short-term trend of decreasing prices and easing off of the activities when prices were rising something that we think most program administrators would be tempted to do. The lowest levels of price volatility were achieved with a combination of government and producer funds followed closely by a scenario with no government funding and a producer assessment of about 13 cents. Importantly, the program functions better with dairy producers having some skin in the game. The best case scenario indicated an average government payment of $900,000 per month (considerably less than the $20,000,000 available in program savings from halting the MILC and DEIP programs) and just a bit less than an 11 cent per cwt assessment on producer milk. The program rule that provided the least volatility was a Program Activation Price (PAP) of $1.44 per cwt. above the average 2004 price level of $16.42 per cwt. That is, expenditures under the DFWT program would be activated if the milk price fell below levels that were $1.44 above the 2004 milk prices. This is not a price-increase target, but a decision rule for the program. That is, program implementation with available funds

22 does not increase the all-milk price by $1.44. The program was more aggressive in export subsidies and herd retirements the further the milk price fell below the PAP. Given this level of the PAP, it is important to note that one of the factors contributing to price stability is the continuous operation of the program. It is always active slaughtering cows and exporting dairy products, and this continuity is important to achieve greater price stability. We think that the DFWT proposal will raise important issues with regard to WTO commitments. Export subsidies impact world market prices and are something that the WTO has explicitly worked to diminish. Our commitments to limited export subsidies for dairy products are a central element of our WTO commitments. The DFWT program assumes that our current Dairy Export Incentive Program ceases as soon as the new program begins and the optimal exports under the new program are not as large as previously. Actually, because of higher domestic prices for dairy products, net imports of all dairy products increase in all scenarios considered. The DFWT program, operating as a government-mandated and possibly funded program, is certain to raise an outcry among other dairy exporting countries. (In particular, this would stand in stark contrast to the marked reduction in export subsidies in recent years by the European Union.) The current CWT program is able to subsidize exports because it is a voluntary producer-funded and limited-scale effort. If federal policy steps in and makes the producer participation mandatory and/or contributes taxpayer dollars to the program, then it seems likely that a number of countries would begin legal proceedings against the U.S. under the WTO. There are positive attributes to the program beyond enhanced milk prices and lower volatility. Cull cow prices do not decrease as producers reduce voluntary culling to produce more milk. This might garner support or at least not opposition from beef producers. Also, unlike the refundable assessment program, producers have a relatively small up-front investment in this program (10-15 cents per cwt versus $1) and individual producers are not penalized for significant growth of their businesses at any time. It may also be the case that more dairy farmers who are at a life stage where they are ready to retire, might be enticed into doing so more quickly or have a more comfortable retirement with an accepted herd retirement bid. Milk price volatility, predictability and adequacy are problems in the dairy industry today. Policy can provide tools to ease these problems and give a more stable environment for rural America. The DFWT proposal appears to be one of those tools

23 Further Steps in the Analysis of a DFWT Program Economic modeling, such as our foregoing analysis, is just what the term implies a simplification of reality. That shouldn t suggest that the results are invalid but rather that such an effort should be regarded as a proof of concept. The analyses account for many of the dimensions of the proposed DFWT program, and suggest that under most conditions, such a program can be successful in reducing price variability and increasing both average all-milk prices and dairy farmer revenues. However, there are a number of characteristics of dairy markets that have not been fully represented in the model used to assess the program. The most important of these include seasonality of milk production and dairy product demand (that is, changes from month to month) and regional differences in utilization and the extent and type of milk market regulation (FMMOs, California, state regulation and unregulated areas). The model used is a national model and thus does not examine regional effects. Important questions can be raised about regional impacts of herd reductions that could be better addressed with a more spatially disaggregated modeling effort. In addition, our analyses focus on underlying (or endogenous) price variation arising from decisions by U.S. dairy producers and processors, not on supply shocks (such as ethanol production) or world market developments (such as our recent exports of nonfat dry milk). Shocks are an important source of price variation in US dairy markets. The ability of a continuously-operated DFWT program to reduce price variation in the face of these random shocks has not been evaluated as a part of this research. Finally, practical issues in the establishment and administration of the program have not been addressed. The analyses reported herein assume that the program is effectively implemented given the assumed decision rules. Although we believe that these factors either individually or collectively are unlikely to completely undermine the effectiveness of the proposed DFWT program, further evaluation and discussion of these effects (and perhaps further quantitative modeling work to more fully address them) are advisable. Thus, our analyses should be considered as an initial assessment that can be complemented by additional research

24 References Nicholson, C. F. and T. Fiddaman Dairy Policy and Price Volatility. Paper presented at the 10th Annual Workshop for Dairy Economists and Policy Analysts, Memphis, TN, April 23-24, Nicholson, C. F., M. W. Stephenson and A. M. Novakovic Assignment of New Products Under Classified Pricing: A Conceptual Dynamic Model of Class Assignment Outcomes. Department of Applied Economics and Management, Cornell University. [Research Bulletin 04-01] Nicholson, C. F. and P. M. Bishop Intermediate Dairy Product Trade and Domestic Policy Options in the US Dairy Sector. Department of Applied Economics and Management, Cornell University. [Working Paper 06-24] Nicholson, C. F. and M. W. Stephenson Market Impacts of High-Protein Whey Product Promotion. Report to Dairy Management, Inc. Nicholson, C. F. and H. M. Kaiser Dynamic Market Impacts of Generic Dairy Advertising. Journal of Business Research, in press. Pagel, E. J Dynamic Patterns of Change in Structure Under Different Support Policy Regimes: An Examination of US Dairy Farming. M.S. Thesis, Cornell University. Pratt, J. E., P. M. Bishop, E. M. Erba, A. M. Novakovic, and M. W. Stephenson A Description of the Methods and Data Employed in the U.S. Dairy Sector Simulator, Version Department of Agricultural, Resource and Managerial Economics, Cornell University. [Research Bulletin 97-09] Schmit, T. M. and H. M. Kaiser Forecasting Fluid Milk and Cheese Demands for the Next Decade. Journal of Dairy Science, 89: Sterman, J. D Business Dynamics: Systems Thinking and Modeling for a Complex World. Boston: Irwin/McGraw-Hill

25 Appendix Table 1. Cornell Dairy Sector Models Developed Since the mid-1990s Model Description Purpose Base Year USDSS (Pratt et al., 1997) Dairy Price Volatility Model (Nicholson and Fiddaman, 2003) Static spatially disasggregated model of US dairy sector with five product categories and two components. No policy elements included. LP model that minimizes assembly, processing and distribution costs. National dynamic (monthly) dairy sector model with detailed product disaggregation, detailed representation of FMMO and DPSP policies. No imports or exports. Identification of normative location values for milk (e.g., indications of location differentials). Has also been used to assess impacts of specific plant closures and changes in location of milk supplies since Analysis of the origins of price volatility in the US dairy industry and how various policy options would influence price variation. 1995; selected updates to 2001 and Dynamic Dairy National dynamic (monthly) dairy Analysis of the impacts of the classification 2001 Product Classification sector model with four products of a new dairy product in Class I Model (Nicholson et al., 2004) and detailed representation of FMMO pricing. No imports, exports or DPSP. Simplified milk supply response. or Class II on dairy farmer revenues. Dairy Farm Structural Change Simulator (originally developed by Pagel, 2005) CHUNK (Nicholson and Bishop, 2006) Cornell Dynamic National Dairy Sector Model (Nicholson and Stephenson, 2006) National dynamic (annual) dairy sector model with detailed representation of farm accounting for four farm size categories and linkages with national dairy markets. Two product categories, simplified representation of Federal Orders, DPSP, DTP, MDP and MILC. No imports or exports. Static market equilibrium model with two US regions, detailed domestic and trade policy representation, detailed product disaggregation and three components. Formulated as Mixed Complementarity Program (MCP). National dynamic (monthly) dairy sector model with detailed product disaggregation, detailed representation of FMMO, DPSP, trade and MILC policies, explicit component balance, detailed dairy product imports and exports. Assessment of how the DPSP influenced changes in dairy farm structure (number and size of farms) over the period Has also been used to assess selected options of the current Farm Bill, and was the basis for the analysis of Refundable Assessments (given the need to model farm participation decisions). Analysis of the impacts of milk protein 2001 concentrate imports and policy alternatives to address them. Has also been used to assess impacts of the Australia-US free trade agreement and selected changes to classified pricing formulas. Model originally developed to assess dynamic market impacts of increases in whey protein product sales. Has also been used to assess dynamic effects of generic advertising expenditures for fluid milk and cheese and to assess the dynamic impacts of changes in make allowances data are used for most analyses