A Regional Economic Analysis of Dairy Compacts:
Implications for Missouri Dairy Producers

In this section the methodology used to study the impact of dairy compacts is described. In particular, a detailed discussion of the economic model developed for this study and the assumptions used in the analysis is provided.

A review of previous studies that investigated the economics of dairy compacts prior to this study reveals that very few studies have been conducted. The Compact Commission in the Northeast has contracted with the University of Vermont to analyze the market impacts of the Northeast Interstate Dairy Compact in the region. At press time, this study is not yet available.

The only published study found on the subject of dairy compacts was "The Economic Effects of the Northeast Interstate Dairy Compact" produced by the Office of Management and Budget (OMB). OMB’s objective was to present an objective evaluation of the economic issues raised by the Northeast Compact: its effects on dairy producers and on their products, customers and competitors. More specifically, the function of the OMB study was:

The following conclusions were gathered from the OMB study:

• New England’s dairy industry is much like that of the rest of the Nation.
• The Compact increased producer prices for New England dairies by $3 per cwt or 25 cents per gallon in July 1997. Prices nationwide began to increase in the last half of 1997. Thus, the difference between Compact prices and New England federal order prices narrowed to just 8 cents or less per gallon.
• The estimated impact of the Compact on raising retail prices ranged from $0.05 to $0.20 per gallon, based on methodology.
• New England dairy farm income rose by an estimated $22-27 million between July and December 1997 (or about 6-8 percent). In addition, milk production rose about 3 percent and milk shipments into New England rose 8 percent.
• Fluid milk sales in New England during July - December 1997 were down 0.7 percent relative to the same period a year earlier. Nationally, fluid milk sales were up 0.2 percent during this same period.
• The Compact’s price regulation reduced the all-milk price to all producers outside New England by an estimated $0.02 per cwt for the period July - December 1997. Gross income for these producers was thus reduced $10-$12 million in 1997.
• Some producers in neighboring areas of the Compact (i.e. New York) may have received higher milk prices as processors outside the Compact region compete with the Compact’s higher price.
• If Congress approves, the Compact could be expanded from 6 up to as many as 12 States (New England plus Delaware, Maryland, New Jersey, New York, Pennsylvania, and Virginia). These 12 states would account for 20 percent of the Nation’s milk supply, compared with New England’s 3 percent.
• If the Compact were expanded to these 12 states and if the price effects were similar to those observed in 1997, the all-milk price to producers outside the Compact region would decline by an estimated $0.10 per cwt. Expanding the Compact just to New York would cause the national all-milk price to decline by an estimated $0.04 per cwt per year.

The objective of this study is to evaluate the economic impact of Missouri joining a Southern Dairy Compact. Changes in Missouri milk production, consumption and farm to retail prices were evaluated.

Missouri cannot have a dairy compact on its own. Therefore it would seem logical to focus the study on just the economics of a Southern Dairy Compact. Even in this instance, one cannot divorce those states involved in a Southern Compact from the rest of the United States because of two reasons. First, raw milk and processed dairy products would continue to be exported and imported into and out of the new dairy compact. Second, prices for manufactured dairy products are determined at the national level. Thus any changes in production and consumption in the Compact states will have ramifications on non-compact states.

For these reasons, a national study was developed as well as a multi-region dairy model that will simulate supply, demand and prices under federal and state orders. One or more dairy compacts were developed and imposed on the federal order structure. This will show the economic impact of compacts on a regional basis. Also state submodels were developed to show what impact a Southern Dairy Compact would have on Missouri.

Conceptualization Issues

Compacts are legally defined as a grouping of two or more states. Thus they are regulated within the strict confines of state borders. The Northeast Interstate Dairy Compact, for example, works within the borders of its member states. The Compact Commission must regulate only fluid milk sales within the borders of these six New England states, and must disburse compact funds only to dairy producers that market milk inside the Compact region. Thus some dairy producers are identified as residing outside the borders of the Compact region. And, pricing decisions are augmented by the New England federal order, which is much smaller than the boundaries of the Northeast Compact.

The point is, compacts work by using state borders. But in reality, most milk in the U.S. is marketed within in the confines of federal orders. One needs to consider this when conceptualizing how a new regional dairy compact may work and how an economic model will be constructed to analyze compacts. For example, there will be two orders that will cover Missouri under proposed federal order reform (Central States order and the Southeast order). The Compact Commission that will be constructed to run the compact must account for that portion of milk that is within the borders of Missouri but is also inside the Central States order. Also, the economic model developed to analyze compacts must use the federal order data and make modifications to account for state borders.

In addition, as mentioned in the OMB study, the benefits of a dairy compact may spill into bordering states as processors in these states compete with higher milk prices in the compact region. That’s because dairy compacts cannot discriminate against outside milk. Some dairy producers in New York are receiving compact prices for milk they ship into New England even though they are producing milk outside the Compact region.

The point is, when considering the formation of a new Southern Dairy Compact, one needs to not only consider which states are going to join, but also which federal and state orders will be affected. In addition, one should evaluate whether dairy producers in bordering states will take advantage of the higher milk prices in the Compact region. In addition, some cities in compact states actually border two states, such as St. Louis and Kansas City. What happens with milk pricing in these two cities if say Missouri joins a new Southern Dairy Compact, but Illinois and Kansas don’t?

To carry out this objective, an intermediate-run regional economic simulation model of the U.S. dairy industry was developed. This model reflects the economics of federal and state orders, and includes a consistent set of elasticities derived from the literature. The model was then "calibrated" to a baseline reflecting regional supply, demand and prices. A number of dairy compacts (i.e. Northern, Mid-Atlantic, and Southeast) were imposed on the model, which in turn was simultaneously resolved. The difference between the baseline and the model simulations would then represent the economic impacts of Dairy Compacts.

Literature

There are a number of economic models available in the literature that reflect the U.S. dairy industry. The Food and Agricultural Policy Research Institute (FAPRI) developed a linear dynamic econometric model of the U.S. dairy industry used to analyze dairy policy options. This model is multi regional on the supply side, reflects farm, wholesale and retail levels, and solves simultaneously for retail prices for butter, nonfat dry milk and cheese. Federal milk marketing orders are approximated in the model by reflecting discriminatory pricing in the computation of farm level milk prices.

Kaiser developed an econometric model of the U.S. dairy industry to evaluate the impact of national generic dairy advertising. This model was similar in structure to an earlier industry model developed by Liu et al. (1990, 1991). It was a partial equilibrium model of the domestic dairy sector (with no trade). It divided the dairy industry into retail, wholesale and farm markets. The model disaggregated manufactured dairy products into frozen products, cheese and butter. All equations were specified in a double-logarithm functional form and were estimated simultaneously using two-stage least squares.

Suzuki and Kaiser estimated an imperfect competition model of the U.S. dairy industry to determine whether the assumption of perfect competition in the U.S. dairy industry biases the findings of economic impacts of generic dairy advertising in the United States. This model also had a double log specification and reflected milk supply, fluid demand and manufacturing demand. The model was estimated using two-stage least squares and solved simultaneously for the manufacturing milk price.

Cox and others developed an interregional competition model of the U.S. dairy industry. Denoted the UW Dairy IRCM, this model was designed to evaluate the effects of specified changes in factors that affect milk and dairy product supply and demand on regional prices, production, consumption and trade flows. In particular, the model was developed to address issues associated with milk pricing under federal and California milk marketing orders.

Economic Model

The federal order model used in this study is based on a discriminatory pricing model of the U.S. dairy industry reflected in a study by Ippolito and Masson (1978). This model described federal milk marketing orders by establishing alternative prices for a single grade of milk in order to maximize farm level profits. More specifically, the model sets a high price for milk used for fluid purposes. The demand curve in this market is characterized as being relatively inelastic.

A lower price is established for milk used for manufacturing purposes. The demand curve in this market is relatively more elastic than that in the fluid market and is more responsive to changes in the price of milk. Thus milk not used in the higher valued fluid market will be attracted to the manufacturing market. Farmers then receive a blend of the prices of milk in the fluid market and manufactured market. In theory, use of discriminatory pricing results in a higher farm milk price than what would occur in an unregulated market.

In this study a standard supply and demand model for analyzing dairy policy was developed that is similar to earlier work by Gardner. A simple supply and demand model was specified using a constant elasticity functional form. Note that this model was not derived using econometric estimation methods like that of Kaiser who used a log linear specification. Rather, a set of elasticities was derived based on a review of the literature and an understanding of the structure of the U.S. dairy industry.

This model can be described as a static equilibrium model that reflects intermediate-run adjustments in the milk supply. The milk supply is specified as a function of the blend price under federal and state milk marketing orders. Raw milk is then allocated according to class uses under federal orders (i.e. milk used for fluid and manufacturing purposes). The amount of milk used for manufacturing purposes then determines the supply of dairy commodities (i.e. cheese, butter and nonfat dry milk) that enters the wholesale market. Demand functions for these products are specified in relation to wholesale product prices. The model then solves simultaneously for the wholesale prices of butter, nonfat dry milk and cheese. That in turn derives class prices and drives milk supply and allocation.

The model was specified to reflect the operations of federal and state milk marketing orders. This was done so that the impacts of creating one or more regional dairy compacts could be simulated. Thus milk marketings and use were specified for each federal order, plus California. For reasons explained later, this model reflects the federal order structure proposed by Secretary Glickman. This new proposal reduced the number of federal orders, introduced new formula’s for class prices and uses a different set of Class I differentials.

This Dairy Compact Model is specified in equations 1- 54 that follows. Marketings and milk use are specified in equations 1-8. Milk marketings reflect the amount of milk marketed in federal order i. Note that this is different than describing milk production within a geographic area. Marketings are specified as a function of the farm price of milk within the order. Thus milk marketings vary with changes in both the farm price and the magnitude of the supply elasticity a .

Milk marketings are then allocated to alternative class uses (classes 1-4). Under federal orders, milk handlers (processors) must pay for milk depending on how it is used. There are different prices for each class use.

Equation 2 specified Class 1 use under federal orders as equal to retail demand for fluid use. Equation 3 is Class 2 use for soft manufactured dairy products. Both equations are inversely related to the respective class prices. For example, the higher the Class 1 price, the less milk is used for fluid purposes. Likewise, the higher the Class 2 price, the less milk is used for Class 2 purposes. This specification is similar to Ippolito and Massen’s specification of milk used for fluid and manufacturing purposes.

Equations 4-7 describe how Class 3 milk is specified. This is a unique specification. It is observed that milk is first allocated to Class 1 and 2 use. The balance of milk (marketings less milk used for Class 1 and 2) is allocated to the residual use for milk: Class 3 for milk used for cheese production and Class 4 for milk used for butter/nonfat dry milk production. One can argue that federal orders were designed to do this. Thus equations 4-7 take residual milk, which is defined as marketings less milk used for Class 1 and 2, and allocates this milk to Class 3 and 4 on a proportionate basis.

The proportion of residual milk used for Class 3 use (Equation 7) is defined as a function of the relative earnings of milk used in either a cheese plant or a butter/nonfat dry milk plant. These earnings (Equations 4 and 5) are simply yields times product prices.

Suppose cheese prices rise. This results in higher farm gate prices. That in turn will increase milk marketings. This increase will flow into both Class 3 and 4 use. A higher proportion of this extra milk, however, will flow into Class 3 use because gross earnings to a cheese plant will rise relative to that of a butter/nonfat dry milk plant.

Equation 8 is a residual identity that defines Class 4 use for milk and clears the market for raw milk.

Class prices in this dairy model are as defined under Glickman’s federal order proposal and are defined in Equations 9-12. These are the prices specified in the milk use equations. Class prices under this proposal are a function of skim milk and butterfat prices. These prices are specified in Equations 25-39. Ultimately all class prices under Glickman’s proposal are specified as a function of wholesale prices for butter, nonfat dry milk and cheese. A formal explanation of class prices under Glickman’s proposal may be found in Bailey (1998) and Cropp and Gould.

Equations 14-18 describe retail fluid milk consumption. Under this model, Class 1 use under federal orders is derived by fluid consumption at the retail level. In equation14, per capita fluid milk consumption is a function of retail milk prices. Equations 15 and 16 describe the linkage between wholesale prices (Class 1 prices under federal orders) and retail prices. Included in this linkage are premiums and the farm-to-retail mark up. Equations 17 and 18 describe total fluid milk consumption and retail fluid milk expenditures.

The farm price of milk is defined in equation 13. It is equal to the sum of class prices weighted by milk use for each federal order. It also includes the value of any over-order premiums on fluid milk sales.

California has a unique state order that is modeled in Equations 40-54. These equations specify class prices unique to the California order. Equations 1-8 were used to specify California milk marketings and use.

All of the milk used for Class 3 and 4 are summed across all federal orders and are then converted into pounds of butter, nonfat dry milk and cheese using Equations 19-21 and milk equivalent conversion factors found in Bailey (1997). Milk produced in California and other regions of the U.S. not proposed to be regulated by federal orders in this computation is included. This then specifies the supply side of the dairy products markets for butter, nonfat dry milk and cheese.

The final set of equations specifies the U.S. demand for butter, nonfat dry milk and cheese in Equations 22-24. These specifications are a function of the wholesale price of each commodity only. All other variables that affect demand (i.e. competing prices, income, tastes and preferences) are implicitly reflected in the intercept terms and do not vary under model simulation.

The Dairy Compact Model solves simultaneously for three wholesale prices: butter, nonfat dry milk, and cheese. Essentially the model attempts to find a price that will set supply equal to demand for each of these dairy commodities. Any changes in those prices affects class prices, which in turn affects milk marketings and use. Changes in both class prices and milk use will also change the blend price at the farm level. That in turn will affect the level of milk marketings. Marketings and milk use are in fact simultaneously determined since milk use alters the blend price. These changes will alter Class 3 and 4 use, which will affect dairy commodity production, which will result in a new equilibrium price for dairy commodities.

The economic model presented here is dependent on a predetermined set of elasticities. Literature was reviewed in order to arrive at a consistent set of elasticities that could be used in this model. The model begins with a review of elasticity estimates for fluid milk consumption and then review those for dairy industry models.

The initial impression of recent studies that derived fluid milk elasticities is that these estimates are much higher than prior ones. Maynard developed an interesting paper that assessed the validity of the argument that consumers benefit from more stable retail milk prices. He estimated a number of models that included price volatility in the demand for fluid milk. His results were interesting in that all elasticities were negative and inelastic, but were at levels five times or greater than prior estimates. His elasticity estimates were: -0.52 to -0.58 for whole milk; -0.33 to -0.72 for 2percent milk; -0.54 to -0.74 for 1percent milk; and -0.61 to -0.81 for skim milk.¹⁴

Gould investigated factors affecting the demand for reduced fat milk. He used household panel data which included over 4,300 households who recorded fluid milk purchased for at-home consumption over a 12-month period. Own- and cross-price elasticities were estimated along with effects of household demographic characteristics. A three-equation fluid milk demand system was estimated for fluid milk that varied by fat content: whole, 2percent and other reduced-fat milks. An econometric model was estimated using the maximum likelihood module, MAXLIK, within the GAUSS software package. The estimated short-run own price elasticities were -0.803 for whole milk, -0.512 for 2 percent milk and -0.593 for 1 percent and skim milk.

AC Nielsen estimated fluid milk elasticities using retail scanner data (Hall). This study, "DMI Milk Pricing Analysis," was conducted under contract with Dairy Marketing, Inc. (DMI). It was designed to help DMI gain a better understanding of how changes in the regular price of milk affect sales of the milk category. Sixteen refrigerated milk items in AC Nielsen Scantrack^â markets over 104 weeks ending August 16, 1997 were analyzed. The analysis was performed using non-promoted store weeks, focused on the relationship between everyday shelf price and sales rate and included three types of analysis: sales rate by everyday price, sales rate by price differential and multi-competitor pricing model. Their findings indicate that a 5 percent increase across all gallon and half gallon milk items resulted in a 1.6 to 3.8 percent decrease in category sales. This would imply an own-price elasticity of -0.32 to -0.76.

The Food and Agricultural Policy Research Institute estimated a structural econometric model used for simulating changes in dairy policy. The econometric model is based on linear specifications using annual time series data. Model parameters were estimated for fluid milk consumption using annual per capita consumption data for whole and lowfat milk as a function of the U.S. average retail price of whole milk. Elasticities were computed using the estimated own price coefficients and the mean of their consumption and price date samples. The results were -0.229 for whole milk and -0.101 for low fat milk.

There are a number of academic studies that estimated demand systems as part of a total dairy industry model. Liu et al. estimated a two-regime dairy structural system. Their econometric model consisted of farm, wholesale and retail levels. They used a double log specification. Therefore the coefficient for the retail fluid milk price index in the per capita fluid milk demand equation (-0.282) is the same as the elasticity.

Kaiser (1997) and Suzuki and Kaiser (1997) estimated dairy industry models to estimate the economic impact of generic dairy advertising. Both models estimated farm milk supply, Class I demand and manufacturing demand using two-stage least squares and quarterly data. Kaiser broke manufacturing demand into cheese, butter and frozen desserts. All equations in these two models were specified in double-logarithm functional form. Suzuki and Kaiser reported own price elasticities of -0.158 for Class I demand, -0.217 for manufacturing demand and a long-run milk supply elasticity of 0.591.

Cox et al. developed an interregional competition model of the U.S. dairy industry denoted the UW Dairy IRCM (Cox (1998), Cox et al. (1995), and Cox et al. (1994)). The model was developed by several researchers in the Department of Agricultural and Applied Economics at the University of Wisconsin-Madison to evaluate alternative dairy policy options, particularly changes in milk pricing under federal and California milk marketing orders. The model is specified for multiple regions that represent separate milk and dairy product production/consumption areas that correspond to federal and state marketing orders. Each region has demand relationships for nine dairy products. The model allocates farm milk supply to various regions and meets local consumption through either local production or imports. The supply and demand elasticities used in the model are intermediate-run.

Heien and Wessells estimated the structure of dairy product demand using Household Food Consumption Survey data. They estimated a complete demand system for food incorporating price and income effects, as well as demographic effects. An almost ideal demand system (AIDS) was chosen to model consumer preferences.

Huang estimated a disaggregate U.S. food demand system. The model has a complete set of food demand relationships consisting of direct- and cross-price elasticities, and expenditure elasticities. An ordinary demand system was derived from first-order differential approximation of conceptual demand relationships. The estimation procedure used constrained maximum likelihood with a substitution approach. The model estimated a U.S. food demand system of 39 food categories, including dairy, and one nonfood sector using annual data from 1953 to 1990. Huang’s analysis included direct price elasticities for cheese, fluid milk, evaporated and dry milk, butter, margarine, and ice cream and other frozen dairy products.

Finally the Office of Management and Budget (OMB) developed a standard economic model used to estimate the price, income, production, and consumption effects of the Northeast Interstate Dairy Compact. The model used milk production and consumption elasticities based on previous USDA econometric studies and was consistent with baseline data for 1997 used in preparing the 1999 President’s Budget. The model assumed that the demand for fluid milk is more inelastic than that for milk used in manufactured dairy products, and that marketing margins are not affected by the Compact’s price regulation.

Results of this intensive literature review of dairy industry elasticities are summarized in Table 4.1 and 4.2 below. The results for fluid milk are interesting in that elasticities derived from annual time series data using aggregated data across many individual markets suggest elasticity estimates much lower than those derived from actual individual sales data. For example, the elasticity range estimated in the AC Neilsen study using actual scan data is in line with elasticity estimates provided by both Maynard and Gould (Table 4.1).

Table 4.1 Own Price Elasticities for Retail Fluid Milk

Note: see the bibliography for references.

Table 4.2 Own Price Elasticities from U.S. Dairy Industry Models

Note: FAPRI = Food and Agricultural Policy Research Institute; OMB = Office of Management and Budget. See the bibliography for references.
¹Long-run elasticity.
²Intermediate-run elasticity.
³Short-run elasticity.
⁴American cheese.
⁵This is not a price elasticity. It applies to the proportion of excess milk used for Class 3 use. See the model specification.

The elasticities used in this dairy compact study are listed in Table 4.2. A milk supply elasticity of 0.35 was used to reflect an intermediate-run response to the farm price of milk. This number was higher than FAPRI and OMB’s short-run elasticity, but lower than Suzuki and Kaiser’s long-run estimate.

The retail elasticity for fluid milk in this dairy compact model is -0.32. This number represents the low end of the elasticity range reported in the AC Nielsen study. It is considerably lower than estimates reported by Maynard and Gould, but three times higher than those reported by FAPRI and Cox. The elasticity of -0.32 was selected since the work of AC Nielsen, Maynard and Gould was based on actual individual sales data and was more recent than some of the other studies. The lower range of the Nielsen study was chosen in order to be conservative.

The elasticity of milk used for Class 2 products in the dairy compact model was adopted from the OMB study. The specification for milk used for Class 3 purposes in the dairy compact study is reviewed in Equation 7. It is a new specification and the elasticity used cannot be compared to the literature.

The rest of the elasticities used in the dairy compact model--those for wholesale dairy product demand--increase in magnitude from cheese, to butter to nonfat dry milk. The reason is that the model of discriminatory pricing that is the basis for federal orders assumes that the own price elasticity for retail demand for fluid milk products is lower than that for surplus dairy products (i.e. manufactured dairy products). Today, milk not used for fluid purposes is manufactured into cheese and butter. "Surplus milk" is used to manufacture nonfat dry milk. Thus the elasticity for nonfat dry milk (-0.60) is higher than that for butter (-0.50) and cheese (-0.35). These elasticities are well within the bounds of estimates from prior academic studies.

Federal Order Reform

A major issue that could affect the results of this study was to decide how to construct a baseline. A baseline is the foundation from which one can analyze the market effects of a dairy compact. The issue was whether to reflect current milk marketing orders, or the proposed orders suggested by Secretary Glickman. The geographic differences between current and proposed orders can be seen in Figures 4.1 and 4.2 below.

Figure 4.1 Marketing Areas Under Federal Milk Marketing Orders as of October 1, 1997

Figure 4.2 Proposed Federal Milk Marketing Orders

Source: http://www.ams.usda.gov/dairy/reform/proprule.htm

This issue could have a significant impact on this analysis of Missouri. Currently Missouri is overlapped by four federal orders: Iowa (no. 79), Southern Illinois-Eastern Missouri (no. 32), Greater Kansas City (no. 64), and Southwest Plains (no. 106). Under Secretary Glickman’s proposed order consolidation, Missouri would have just two orders: the Central order which would cover the northern portion of the state, and the Southeast, which would cover the southern portion of the state. Central Missouri would still be unregulated.

Missouri’s farm gate milk prices could be significantly affected by order consolidation due to changes in Class 1 differentials and percent Class 1 use. Thus, in order to get a clear idea of how a dairy compact would affect Missouri, it was decided that the baseline constructed for this analysis would reflect Secretary Glickman’s proposed federal order structure.¹⁵

Secretary Glickman's proposal can be summarized as follows:

• Have fewer orders (down from 31 to 11 orders, not including California);
• Replace the basic formula price (BFP) with a multiple component pricing system;
• Make fluid milk prices (Class I) more market-oriented over time, yet more stable;
• Make minor revisions to products included in each class of milk;
• Have common provisions to streamline orders;
• Put in place a transition program to help farmers become more market-oriented.

The regions in Table 4.3 describe the new proposed federal milk marketing orders.

These orders represent 11 of the 13 regions developed in this Dairy Compact Model (the other regions are California and unregulated areas). For a graphical representation of the proposed federal order regions, see Figure 4.2 above.

The current and proposed Class 1 differentials are detailed in Table 4.4. Secretary Glickman proposed a whole host of options for Class 1 differentials. The least controversial option was included: Option 1A. This option is most similar to current Class 1 differentials.

Note that the current orders are aggregated as subheadings under the proposed new orders in Table 4.4. In addition, the annual average market over-order premiums for Class 1 milk sales in 1997 are listed. This information will be important later when discussing the baseline.

In addition to new orders and Class 1 differentials, Secretary Glickman’s proposed order reform will also result in new definitions of class prices. These new changes are outlined in detail in Equations 25-39 and in Bailey (1997).

Construction of a Baseline

The analysis of dairy compacts begins with the construction of a "reasonable" baseline. A baseline is a starting point from which to estimate changes in supply, demand and prices due to the creation of a dairy compact. Most economic studies select a previous year from which to construct a baseline using actual data. Such an approach can be called "reasonable" since actual data is being used.

The baseline used in this study reflects the proposed changes in federal milk marketing orders. These proposed changes have not yet been adopted. Thus, a full year of data for supply, demand and prices that reflects these new changes is not available. A "reasonable" baseline that incorporates all of the proposed changes in federal orders discussed so far has to be created.

The methodology for constructing a baseline from which to conduct this study of dairy compacts can be outlined as follows,

1. Create a baseline for 1997 that reflects milk marketings, class use and class prices for 31 federal orders and California.
2. Forecast milk marketings and class use for each of the 31 federal orders and California, and dairy commodity supply, use and prices for 1999.¹⁶
3. Aggregate the 31 federal orders into the 11 proposed federal orders.
4. Create formulas for the new proposed class prices and link them to dairy commodity prices.
5. Incorporate the elasticities and model equations outlined above into the model and line up all intercepts to the new baseline.

Table 4.4. Class I Differentials and Market Over Order Premiums in 1997

The first step was to develop a baseline for 1997 using federal order data. The purpose for developing this detailed baseline was to create a reasonable foundation from which to create a new baseline that reflects Secretary Glickman’s proposed changes in federal orders.

The objective in creating the 1997 federal order baseline was to develop a consistent set of data for each federal order that reflected:

1. Milk marketings
2. Class use (I, II, III, and IIIa) and percent use
3. Class prices (I, II, III or BFP, and IIIa)
4. Commodity supply, demand and wholesale prices (cheese, butter and nonfat dry milk).

The data set constructed for 1997 is in Appendix Table 1. The major source of data for this baseline was the Agricultural Marketing Service of USDA (April 1998, June 1998). Additional sources were provided by annual summaries and informal reports provided by individual market administrators. The data for California was provided by the California Department of Food and Agriculture.

One problem in creating this 1997 baseline was that Class IIIa use (milk used for the production of nonfat dry milk) was restricted in some federal orders. USDA combined data for Class III and IIIa milk use whenever there was just one reporting plant producing nonfat dry milk within an order. That created a problem for this model’s baseline since Class III and Class IIIa use had to be separated. To get around this problem, Class III and IIIa use was individually approximated in some federal orders. Regional sums of Class IIIa use and state production of nonfat dry milk were reviewed. Also, numerous staff at market administrator offices were interviewed to learn which orders that combined Class III and IIIa sales produced the bulk of the nonfat dry milk for the region. Through this iterative process the breakout of Class III and Class IIIa use in restricted orders was estimated.

The foundation for the baseline used in the model was created by developing a forecast for the year 1999. This forecast is also in Appendix Table 1.¹⁷ Milk marketings for each order was a starting point. A percentage change from the 1997 baseline was approximated based on industry forecasts provided by FAPRI and an understanding of the marketplace. Thus the 1997 baseline was merely created as an aide in constructing the 1999 baseline used in this study. Some orders expanded milk marketings (i.e. Northeast, Upper Midwest and West), others declined (i.e. Central States, Southeast, Dallas). Milk was reallocated by assuming Class 1 and 2 use would increase with population (i.e. one percent a year), and the balance of any changes would take place in Class 3 and 4 use. Expanded marketings were not used for Class 4 use if the order did not previously have Class 4 use.

Class prices for 1999 were developed from Equations 25-39 based on dairy commodity market price forecasts. Wholesale price forecasts for 1999 for American cheese, Grade AA butter, and nonfat dry milk taken from the FAPRI baseline was a starting point. Under Secretary Glickman’s proposal a NASS survey was developed to estimate national average wholesale prices. In this study, it was assumed that the NASS butter survey price would be $0.09 per pound below the Grade AA butter price in Chicago; the NASS survey for nonfat dry milk would be a penny below the Central States price for nonfat dry milk; and the NASS cheese price survey would be six cents per pound below the 40-pound block cheese price in Chicago.

Once the 1999 forecast for 31 federal orders was completed, it was aggregated to the new consolidated orders. For most orders that meant a simple addition of marketings and class use. For example, the new Florida order was equal to the old Upper Florida (no. 6), Tampa Bay (no. 12) and Southeastern Florida (no. 13) orders. Other orders were a bit tricky. For example, it was assumed that 40 percent of the Southwest Plains order would fall into the new Central order. In addition, it was assumed that 10 percent of the Southern Illinois-Eastern Missouri order and 60 percent of the Southwest Plains order would fall into the new Southeast order. That’s because most of the milk in the Southwest Plains order is in southwest Missouri and northeast Arkansas.

The aggregated data for Marketings and class use would also require the creation of a weighted average Class 1 differential. This would aggregate the Class 1 differentials in individual orders up to one differential for the new order. It was assumed in this study that option 1A would be ultimately used by the Congress. This would create new Class 1 differentials for every county of the U.S. According to USDA:

In this study a weighted average Class 1 differential for each of the 11 orders was created by weighting the Option 1A Class 1 differentials in each of the city centers for each of the existing 31 federal orders. A weighted average Class 1 differential for each of the new 11 orders was then constructed. This is illustrated in Table 4.5.

State Models

The Dairy Compact Model was specified for federal and state milk marketing orders. In other words, model equations to simulate milk marketings, class use and class prices for each order were developed. Thus any scenarios for creating dairy compacts would be imposed on these model equations of individual orders.

Some of the questions regarding dairy compacts, however, are how dairy producers and consumers within the confines of an individual state are affected. The objective of this report, for example, is to access how Missouri dairy producers and consumers would be affected if the state were to join a Southern Dairy Compact.

Table 4.5 Sum of Federal Order Data Used in the 1999 Baseline

To address these questions, various state submodels were developed. These submodels have some independent equations and equations that are specified as a function of the regional equations. None of the state model equations, however, feed back into the main model. The state models are only used to illustrate how changes in regional federal orders due to the creation of dairy compacts could affect farmers and consumers within a state.

Take Missouri for example. A Missouri submodel was developed that estimates per capita fluid milk consumption, total fluid milk consumption, value of fluid milk consumption, farm milk production and marketings and the value of farm milk sales. Additional state models for New York, Wisconsin, Georgia, California and Kentucky were developed.

Let’s start with Missouri milk sales. Missouri milk marketings were forecast to be 2.225 billion pounds in 1999. The following identity was then specified to simulate the impact of changes in farm prices on the level of milk marketings:

Thus, the higher the effective farm price for Missouri, the more milk is produced and marketed. The magnitude of the supply elasticity and the change in the effective farm price determines how much milk marketings change.

The next step was to specify an effective farm price for Missouri. In this case, information about various federal orders would be needed. In particular, there was a need to know class prices and use for any federal orders that overlap Missouri. Missouri is expected to be overlapped by two federal milk marketing orders: Central States and Southeast. It was assumed class use for any milk marketed in Missouri would be affected by these two orders. In other words, the percent use of Class 1 milk in the Central State and Southeast orders will determine the average Class 1 use for Missouri. To simulate this, there is an identity that states 65 percent of the percent Class 1 use in the Southeast order and 35 percent of the percent Class 1 use in the Central states order will determine the average Class 1 use for fluid milk in Missouri. The same was projected for Class 2, 3, and 4 use. The result is class use of milk marketings in Missouri for the baseline year is 61.9 percent for Class 1 use, 10.8 percent for Class 2 use, 26.0 percent for Class 3 use and 1.3 percent for Class 4 use. Based on these formulas, any change in class use in either of the Central States and/or Southeast orders will alter the class use in Missouri.

The estimates for class use were then used to construct a Missouri average blend price. The Class 1 price for Missouri is equal to the Class 1 mover for all federal orders plus a state-average Class 1 differential of $2 per cwt. The other class prices are the same as defined in other orders. The federal order blend price for Missouri then is computed by taking a weighted average of the Missouri class prices. A Missouri effective farm price then is equal to the Missouri blend price plus any market or compact premiums weighted by Class 1 use. Note that Class 1 premiums have been separated into market premiums, which are negotiated by dairy cooperatives based on market conditions, and compact premiums which are equal to the difference between the compact price and the Class 1 price. Again, it was assumed that these market or compact premiums for Missouri would be enforced in the regional federal order model.

The Missouri model also estimates fluid milk sales and retail fluid milk expenditures. The following equations are used:

Per capita fluid milk consumption is specified in equation 56 above. It is a function of the retail fluid milk price. Given a negative fluid milk elasticity, the higher the retail price, the lower per capita fluid milk consumption.

The retail fluid milk price is a simple identity equal to the cost of Class 1 milk to bottlers, plus the farm-to-retail mark up (Equation 57). Note that this mark up is a dollar mark up as opposed to a percent mark up. Equation 58 defines the Class 1 cost of milk to bottlers. It includes three things: the minimum federal order price for Class 1 milk, market over-order premiums negotiated by dairy cooperatives and any compact premiums. These three prices are then converted from cwts to gallons. Total fluid milk consumption and retail fluid milk expenditures are defined in Equations 59 and 60.

Again, the major reason for the state submodels is to take the regional responses to a dairy compact and convert this information to the state level. Thus regional prices are used in various state submodels to generate farm-level milk sales and retail fluid milk expenditures.

There were a number of assumptions made in this analysis of dairy compacts that have direct bearing on the results presented. Outlined are the critical assumptions and the rationale discussed. The assumptions were laid out to be objective and balanced. In some cases, the model was rerun with alternative assumptions to illustrate how they affected the results. These assumptions were based on input received from various members of the Missouri dairy industry.

Level of Compact Premium

There are two critical issues regarding dairy compact premiums that are of concern in this study. First, whether to use a fixed compact price in the model (i.e. $17 a cwt in the Southeast on all fluid milk sales) or a compact premium (a price wedge, say $2/cwt). In the Northeast Compact, for example, the commission fixes a compact price. That then establishes a price floor for fluid milk. A premium is then realized each month. Second, what assumptions are made regarding market over-order premiums under compacts.

Let’s start with the compact premium issue first. The model was developed so that a dairy compact can be simulated in every federal milk marketing order. For this analysis there is interest in only two compacts: a Northern Dairy Compact, and a joint Mid-Atlantic/Southern dairy compact. But what level of premium should one impose on the model to simulate a compact?

One option would be to fix a compact price, say $17 per cwt. The compact premium would then be defined as the difference between $17 and the Class 1 price in each of the federal orders in the compact region for the baseline year. The problem with this option is that one must choose a justification for the $17 level. Shouldn't the compact price be different in every order since Class 1 differentials vary so much?

Another option would be to use a price wedge. In this case, one would set the compact price so many dollars above the existing Class 1 price. For example, one could choose a compact price in the Northeast and Southeast that is $2 per cwt above the existing Class 1 prices. This option makes better sense for this study since one can target a price change. Thus in this study, all analyses of dairy compacts is based on price wedges.

Note: with an annual economic simulation model, there is no difference between setting a compact price or using a price wedge or premium. The impact on the results is the same! Thus, in the real world, a price floor is set and a monthly compact premium is collected and disbursed to farmers. This monthly compact premium will vary from month to month. The fluid portion of a farmers milk check, however, will be fixed at the price floor. In an annual simulation model, only the annual premium is incorporated into the model and analyzed.

Another issue is what happens with market over-order premiums in those federal orders that have a dairy compact? For example, there is a $1.23 per cwt market over-order premium in the baseline for the Southeast federal order. What assumptions are needed regarding this market premium if a $2 compact premium is imposed on the model? Does one assume the market over-order premium of $1.23 holds when one imposes a compact? Or, does one assume that the market over-order premium falls in half or down to $.045 per cwt, or disappears?

In this study it is assumed that the purpose of a dairy compact is to stabilize and enhance farm level milk prices. A compact price would be set in order to gain a certain price premium. Thus, to identify the economics of dairy compacts and their impacts on the market, this study retains current market over-order premiums set in the baseline and creates a compact premium. How would one analyze a $2 compact premium if market premiums fall $1? Wouldn't that be the economics of a $1 premium? From the model's perspective, a market and compact premium is identical.

Thus all of the analysis that follows will make the assumption that fluid processors will pay the Class 1 price in the order plus a $2 per cwt compact premium. This is similar to the compact premium that existed in the Northeast Dairy Compact during the first six months of operation in 1997. It is also assumed that any market premiums that existed in the baseline would remain in tack. In this way, the economic impact of a $2 compact premium can be isolated. If one wants to reduce market over-order premiums or analyze a $1 compact premium, just divide the results in two.

Number of States in Compacts

The next major issue is to define which dairy compacts will be studied and what the borders of these compacts will look like. There are a number of options. One option is to decide which states to include in which dairy compacts. The issue of forming dairy compacts is currently being decided on a state-by-state basis. Thus for purposes of this study, one or more dairy compacts can be created with select states in each compact.

In the Northeast, Delaware, New Jersey, New York, Pennsylvania, Maryland and Virginia are considering whether to join the existing Northeast Interstate Dairy Compact. Under Senate Joint Resolution No. 28, these states can join, "… if on entry the additional States are contiguous to participating States …". Congress must also consent.

Also, there is an effort currently underway to create a new Southern Dairy Compact. An attempt is being made to pass common legislation in each of the participating statehouses. That is a requirement before passing federal legislation to create a Southern Dairy Compact.

The point is, dairy compacts are being formed on the basis of state boundaries even though the enforcement of these compacts will likely take place within the boundaries of federal orders.

In this study three dairy compact regions were created without regard to state borders. The baseline in this model does not contain any dairy compacts, not even a Northeast Interstate Dairy Compact. Thus alternative scenarios will reflect the impact of one or more dairy compacts. These compacts will not be defined on the basis of state boundaries. This approach was not used since individual federal orders were modeled. Thus any dairy compacts analyzed will for the most part be within the geographic boundaries that apply to specific federal orders.

The next question then is which federal orders should one include in each dairy compact region? Another question is what to do about surrounding states and unregulated regions. Since farmers in adjacent orders can ship milk into dairy compact regions and receive the higher price, shouldn’t this be reflected in the analysis?

To illustrate how new dairy compact regions will be created in this model, let’s take the case of the existing Northeast Interstate Dairy Compact. That compact was entered into by Vermont, New Hampshire, Maine, Connecticut, Rhode Island and Massachusetts. Practically speaking, this includes New England order no. 1, the northern portion of Vermont and New Hampshire, and all of Maine.

Under federal order reform, a new Northeast federal order will be formed which will include the old New England order no. 1, the old New York-New Jersey order no. 2, the old Middle Atlantic order no. 4 and currently unregulated portions of northern Vermont, New Hampshire and northern and western New York. This new federal order will still not include the state of Maine, a large portion of central Pennsylvania and the Western tip of New York.

Three new dairy compacts will be created that will be used in this analysis. These compacts will abstract from existing and proposed dairy compacts. Since the objective here is to examine the economics of dairy compacts, this simplification will not significantly affect the results.

The Northern Dairy Compact proposed here is geographically very different from the existing Northeast Interstate Dairy Compact. The proposed Northern Dairy Compact will include the proposed Northeast federal order and the state of Maine. It will look something like an expanded Northeast Interstate Dairy Compact, but will not include central Pennsylvania and Virginia. The assumption here is that most of the milk produced in Pennsylvania will be marketed into either the proposed Mideast or Northeast federal order. Thus, this accounts for the majority of this milk.

Virginia will be accounted for in a Mid-Atlantic Dairy Compact along with the proposed Appalachian order. Also, a Southeast Dairy Compact will be formed for this study that will include both the Southeast and Florida proposed federal orders, and that portion of northern Missouri that will be in the proposed Central federal order. This northern portion of Missouri accounts for about 16.4 percent of the state’s total milk production.¹⁸

One can note that this definition of a Southeast order does not include Oklahoma and Texas, two states that showed interest in a proposed Southern Dairy Compact. The economic impacts of the Southeast dairy compact would be greater if these two states were to be included.

Table 4.6 Regional Boundaries for Dairy Compacts Used in the Model Simulation

An allowance was not made for outside milk to flow into the defined dairy compacts. In other words, the model does not allow for Ohio milk to flow into the Northern Dairy Compact and receive the higher compact prices. In the real world, it is known that some milk produced in New York is allowed to flow into the Northeast Interstate Dairy Compact and receive the higher compact premium. The compact cannot legally erect a barrier to outside milk. Thus compact administrators must first calculate the compact premium for a particular month, and then distribute this premium back to all farmers that participate in the dairy compact (farmers located in the dairy compact and outside farmers marketing milk into the dairy compact).

This model does not explicitly account for milk flows between federal orders. It is assumed that federal orders as defined in this study only reflect milk marketed and pooled inside the federal order. That's because federal milk marketing orders reflect where milk is marketed, not necessarily where it is produced. Thus federal orders in this study reflect milk pooled in the order and does not separate locally produced milk from outside milk production.

It is conceivable, however, that creation of a dairy compact would in reality attract more outside milk into a federal order pool located within a dairy compact region. In this case less milk would be pooled in the non-compact orders and more in the compact orders than is reflected in this study. In that case, some farmers outside the compact region would benefit. Also, the higher the compact premium, the more milk will be shifted from one federal order to another. In any event, this is not expected to be a significant volume of milk and will not likely bias the results in a statistical sense. However, since this study did not explicitly allow for inter-regional shifts in milk marketings, the model results could be greater than what is portrayed in this study!

Farm to Retail Milk Margins

Another assumption made in this study is that the farm-to-retail price margin for fluid milk is a fixed dollar amount. This was computed for the baseline year of 1999 by subtracting the Class 1 cost of milk from the retail price of fluid milk. Retail fluid milk prices from AC Nielsen from 1997 was used in the 1999 model baseline.

There are two basic ways that retail prices are established. Both approaches start with the Class 1 cost of fluid milk which includes and Class 1 premiums. One way is to use a fixed percentage mark up. For example, if a fluid processor/retailer pays $1.25 per gallon for raw milk, which represents the farm price of milk, a 100 percent fixed percentage mark up would result in a $2.50 retail price ($1.25*(1+1.00)). If the wholesale Class 1 cost of milk were to rise to say $1.40 per gallon, the retail price would rise to $2.80 ($1.40*(1+1.00)). Thus the farm-to-retail margin increases in proportion to the cost of milk.

Another approach is to use a fixed dollar mark up. Using another example, suppose the cost of milk to a fluid processor/retailer is $1.25 and that the industry standard for the farm-to-retail margin is $1.25. That would result in a retail price of $2.50 per gallon ($1.25+$1.25). The farm-to-retail mark up would still be 100 percent. Now, suppose the cost of Class 1 milk rises to $1.40. Using a fixed dollar mark up, the retail price would rise to $2.65 per gallon ($1.40 + $1.25). The farm-to-retail mark up would now fall to 89.3 percent (2.65/1.40 - 1).

In this study a fixed dollar mark up was assumed for three reasons. First, a fixed percent mark up appears naïve since retailers take other factors into consideration when setting retail prices such as promotions, market competition, seasonality, etc. Second, marketing and operating costs do not depend on the price of milk. Third, this is the assumption used in the OMB study.

It is important to understand that many factors outside of dairy compacts affect retail pricing (i.e. competition, seasonality). However, since this model analyzes changes relative to a baseline, the assumption of a "pass through" of any higher Class 1 costs of fluid milk to the consumer appears to be reasonable. The degree of pass through (i.e. fixed percent or fixed dollar farm-to-retail mark up) is to be debated.

A fixed dollar mark up is a more realistic assumption to use in this analysis rather than a fixed percent mark up. In any event, model results will also be presented for compact scenarios using a fixed percent mark up. Reality may in fact lie somewhere in between.

To reiterate, the results presented in this study are conditioned on the following critical assumptions:

1. A fixed $2 per cwt compact premium used in all of the compact scenarios,
2. Market over-order premiums remain unchanged in the compact scenarios,
3. A fixed dollar mark up will be used in the farm-to-retail margin for fluid milk prices. That dollar margin will remain unchanged under model scenarios.

Again, the model was aligned to a baseline that reflects a projection of supply, demand and prices for federal orders as defined in Secretary Glickman’s proposal. The 1997 federal order data and California data were used as a basis for constructing a baseline for the year 1999. This baseline does not include a Northeast Interstate Dairy Compact. Alternative scenarios were developed to estimate the impact of dairy compacts relative to the model baseline that reflected no compacts. The following scenarios were developed and simulated in the model:

Scenario 1: Northern Dairy Compact. Under this scenario a compact price wedge of $2 per cwt was introduced in the Northeast federal order and in the New York submodel. A $2 per cwt compact premium for Maine was included, which is part of the unregulated region in this baseline model. A weighted average compact premium was used for this region (weighted by the ratio of Maine milk production to milk production in the unregulated region).

Scenario 2: Joint Mid-Atlantic and Southeast Dairy Compact. In this scenario a $2 per cwt compact premium in the Appalachia, Florida and Southeast orders was introduced. Also introduced was a $2 compact premium in the Missouri, Georgia and Kentucky submodels. A compact premium was reflected for Virginia and Northern Missouri. In the current baseline, Virginia is included in the unregulated region and Northern Missouri is in the Central order. Compact premiums are reflected to farmers in these two states by using a weighted compact premium in the unregulated region and in the Central order.

Scenario 3: Combined Northern, Mid-Atlantic and Southeast Dairy Compacts. This scenario reflects all of the changes in Scenarios 1 and 2 above.

How much milk will be accounted for in these compact scenarios? The answer is in Table 4.7 below.

Table 4.7 Percent of Milk in Dairy Compacts

All of the above scenarios assume there is a cost to running a compact. Used here is the same cost structure currently used in the Northeast Interstate Dairy Compact. That Compact charges processors an administrative fee of $0.02 per cwt and holds back a reserve fund fee of $0.04 per cwt on all fluid milk marketed under the compact. Part of the reserve fund fee in the Northeast Interstate Dairy Compact is being held in the event that USDA claims it has incurred higher costs for running the dairy price support program due to the compact. In addition, a 3-percent fee is withheld from the compact premium to offset higher costs for operating the federal Women, Infants and Children’s (WIC) program.

Note, the fees withheld due to higher USDA costs for operating the dairy price support program will be unnecessary after December 31, 1999 since this program is set to expire.

A similar fee structure was reflected in this study. The administrative fee was raised to $0.04 per cwt to reflect the costs of running much larger Compact programs. In this study, it was assumed that this administrative fee is deducted from the compact proceeds. The reserve fund fee was eliminated. The WIC program charge was maintained at 3 percent of the compact premium. Thus fees totaling $0.10 per cwt are deducted from the $2 compact premium in the model scenarios.

The baseline model was developed in a spreadsheet using Microsoft Excel. The model was designed to solve for new prices for butter, nonfat dry milk and cheese after a change is introduced to the system. Pressing the F9 key represents one iteration. Exel’s Scenario Manager was used to document and introduce the scenario changes outlined earlier. Thus this spreadsheet was programmed for four scenarios: the model baseline plus scenarios 1-3 above. Using Scenario Manager allowed us to 1) document all of the scenario changes, and 2) ensure that all of the scenarios represent changes from the same baseline. Interested parties should contact the authors if they wish to have a copy of the Excel Dairy Compact Model.

Interested parties should contact the authors of this study if they wish to learn more about the model.

As stated earlier, some of the model assumptions outlined earlier will be altered and scenarios 1-3 will be rerun to identify the impact these changes have on compact results. For example, the model will be resolved using a fixed percentage farm-to-retail mark up in retail demand equations and will equalize the own-price elasticities used in the Class 1 use and per capita fluid milk consumption equations. These additional simulations will be made in order to quantify the impact the model assumptions make on the results.