Impacts of the Northeast Interstate Dairy Compact on the WIC Program:

Evidence from Boston and Hartford

Qingbin Wang, Zooyob Anne, Catherine Halbrendt, Charles Nicholson, and Jaimie Sung

Department of Community Development and Applied Economics

The University of Vermont

Introduction

The Northeast Interstate Dairy Compact (NIDC) is the first regional dairy compact in the US. It has been the focus of a great deal of attention and speculation in the past several years, especially since its inception in July 1997. Three additional states have been authorized to join the NIDC and more states want to join the NIDC or establish their own regional compacts (Bailey and Gamboa 1999, Knutson 1999). Despite the strong current interest in information or expansion of dairy compacts, there is relatively little published information on the impacts of the one dairy compact already in existence. This study examines the impacts of the NIDC on the Women, Infants and Children (WIC) program using data from Boston in Massachusetts and Hartford in Connecticut.

The mission of the NIDC is to assure the continued viability of dairy farming in the northeast and to provide consumers of an adequate local supply of pure and wholesome milk. To accomplish this, the NIDC has stabilized the price paid by fluid milk processor at $1.40 per gallon since July 1997 through a variable compact order-over premium (see Figure 1). Although the NIDC has helped many farmers in the region, some observers associate it with increases in retail milk prices in the post-Compact period, especially in July 1997. Figure 1 indicates that retail milk prices in Boston and Hartford increased by about $0.20 per gallon at the start of the compact and then remained at levels higher than those observed in previous years. Such increases in retail milk price have brought about concerns from consumers and policymakers. An increase in retail milk price can have a negative effect on consumer welfare and can result in changes in government programs such as the WIC program.

The WIC program was established in September 1972 to prevent anemia and inadequate growth common to children in low-income families. This federal program was designed to improve the health status of participants by (i) assuring access to health care, social and health programs, (ii) teaching families with nutritional practices, and (iii) providing individually-designed nutritious food packages. Eligibility for the WIC program is based on residence, age, maternity status, income, and medical or nutritional need. Fluid milk has been a major food item of the WIC program. According to an estimate of the Food and Nutrition Service of the USDA, WIC programs nationwide spent an average of 30% of their food budgets on fluid milk in 1996.

The NIDC was established as a provision of the 1996 Farm Bill. According to the provision, the NIDC Commission reimburses the WIC programs in the compact states by the amount of the compact over-order premium. This provision was designed to prevent any increase in the net price paid for milk by WIC programs in the compact states. However, because most WIC programs in New England allow participants to purchase milk directly from retail stores, the net price paid by WIC programs depends on both the compact over-order premium and any related changes in the retail price. The net price paid by the WIC program will not be affected only if the change in retail milk price is equal to the compact over-order premium. However, this is unlikely to be observed because changes in the retail price are determined by many factors such as Class I milk price and mark-up behavior of wholesalers and retailers.

Any increase in the cost of fluid milk would increase the food package costs for WIC programs. This could possibly lead to changes such as a reduction in the number of WIC program participants and a change in the mix of food packages. This study does not examine whether the mix of food packages and the nutrient sources have changed. This paper first examines the changes in the number of WIC participants in the compact states and then analyzes the changes in the net milk price paid by WIC programs using data from Boston and Hartford. WIC participants in Massachusetts and Connecticut account for about 67% of the participants in all New England states.

As shown in Table 1, the number of WIC participants in the six states in New England dropped slightly from 257,566 persons in June 1997 to an average of 256,894 in the next three months (July-September 1997), but then increased to an average of 266,844 during October 1997 to February 1998. While the number of WIC participants decreased slightly in Maine, New Hampshire, Rhode Island and Connecticut during July 1997 to February 1998 and increased in Vermont and Massachusetts in the same period, the total number of participants in these six states increased to an average of 263,113 in the post-Compact period (July 1997 to February 1998). Although these changes may be due to other factors, they provide evidence that the price regulation under the NIDC has not had any dramatic effect on the number of WIC participants. Note that this paper does not examine possible changes in the mix of food packages of the WIC programs due to data limitations.

The main purpose of this study is to assess the possible impact of the NIDC on the WIC program. Specifically, we address the following two questions:

· Has there been any significant change in the net milk price paid by WIC programs in the post-Compact period?

· Has there been any significant change in the mark-up behavior of retailers in the post-Compact period that has affected the net price paid by WIC programs?

This study has two principal limitations. First, our analysis is limited to only two New England cities (Boston and Hartford) because retail price data are available for only these two cities. Second, because the NIDC has been in effect since July 1997, the relatively small number of observations for the post-Compact period may limit the analysis of its economic impacts. Readers should interpret and use the results from this study with caution. Further research based on a longer period is needed to more fully examine the impact of the Compact price regulation on the WIC programs.

Methodology

To address the above two questions, we use an econometric technique called an autoregressive-moving average (ARMA) model that depicts the relationship between retail milk price, its past values, past prediction errors, and other factors such as Class I milk price. This section first describes the data and then presents the econometric specification of the ARMA model and the hypothesis to be tested.

Data

Three monthly price series were used in this study: (1) Retail milk prices for Boston and Hartford, (2) Class I milk prices for Zone 21, and (3) Compact price (i.e., Class I milk price plus the compact over-order premium). We used the data from January 1990 to June 1997 for estimating the model and the data from July 1997 to June 1998 for both forecasting and testing. We used the time series data for only seven and half years (90 months) in this study in order to minimize the impacts of structural changes in US dairy markets in the late 1980s. Since the price regulation under the NIDC began in July 1997, the over-order premium set by the Compact Commission is available for July 1997 to June 1998. As discussed above, this study is limited to two New England cities due to data availability.

An ARMA (p, q) Model of the Milk Price System

In this section, we present a procedure to test the impacts of the Compact on the WIC program based on the estimates of an autoregressive-moving average (ARMA) model. In an ARMA model, the dependent variable like retail milk price is a function of both its past values, past errors and current and past values of other time series. The model can be used to examine the relationships and to forecast future values of the dependent variable (Pindyck and Rubinfeld, 1991; Greene, 1993).

In order to investigate the changes in retail milk price under the Compact in Boston and Hartford, we consider the Class I milk price for Zone 21 (P^I), the Compact milk price (P^C), retail prices in Boston (P^B) and Hartford (P^H), and the net price paid by the WIC program (P^W). The relationships of these variables can be expressed as follows:

(1) P^C = P^I + R,

(2) P^j = P^C + m^j for j = B or H, and

(3) P^Wj = P^j – R,

where R is the over-order premium set by the Compact Commission, which did not exist during the pre-Compact period, m^j is the mark-up at market j, which is defined as the difference between the retail price at j and the Compact price in the post-Compact period and the Class I milk price in the pre-Compact period, and R is the reimbursement by the Compact Commission to the WIC program (i.e., the over-order premium). Note that the net price paid by WIC programs in the pre-Compact period was equal to the retail price (i.e., P^Wj = P^j).

Then, the retail price at j is

(4) P^j = P^C + m^j = (P^I + R) + m^j

and the net milk price paid by WIC program at j is

(5) P^Wj = P^j – R = (P^I + R) + m^j – R = P^I + m^j.

Equation (5) indicates that if there is no significant change in the mark-up behavior between the pre-Compact and post-Compact periods, the net milk price paid by WIC programs should not have any significant change. In other words, during the post-Compact period, if the actual retail price minus the reimbursement is not significantly different from the retail price predicted without the Compact, this provides evidence that the Compact did not affect the net price paid by WIC programs.

Therefore, our task is to obtain the predicted retail milk prices and mark-ups that would exist assuming there was no Compact for the post-Compact period. To obtain the predicted retail price, we estimate a time-series model of retail price using information for the pre-Compact period and, based on the estimates, forecast values for the post-Compact period. In the model analyses, we employ the mixed autoregressive-moving average process of order (p, q), ARMA(p, q), of retail price.

The ARMA model of a time-series process, y(t), can be represented by

(6) f(B)y_t = X_tb + q(B)e_t

where f(B) = 1 - f₁B - f₂B² - … - f_pB^p, q(B) = 1 - q₁B - q₂B² - … - q_qB^q, fs and qs are autoregressive and moving-average parameters to be estimated, X_t is a vector of exogenous variables, with or without lags, which affect the process y_t, and b is the corresponding vector of parameters to be estimated. Note that B is the backward shift operator, i.e., B^kz_t = z_t-k.

The procedures to investigate the impact of the Compact on the net milk price paid by WIC programs are as follows:

(i) Estimate the ARMA model for the pre-Compact period;

(ii) Forecast the net retail prices for the post-Compact period, P^j_t*, based on the estimated ARMA model;

(iii) Calculate the net milk price paid by WIC programs over the post-Compact period, P^Wj_t = P_t^j - t_t; and

(iv) Test the hypothesis that the net price paid by WIC programs in the post-Compact period is equal to the predicted price without the Compact by comparing P^W_t and P^j_t* (i.e., the null hypothesis is P^Wj_t = P^j_t* and the alternative is P^Wj_t ¹ P^j_t*)

The increases in retail milk prices since the Compact went into effect in July 1997 are due to two major effects: (a) the effect of increases in the price paid by fluid milk processors through the compact over-order premium, and (b) the effect of changes in the mark-up behavior of wholesalers and retailers. A similar procedure is used to examine changes in the mark-up behavior in order to test whether there was a significant change in the mark-up since the inception of the Compact.

Results and Discussion:

As mentioned above, changes in the retail milk price since July 1997 can be due to two major factors: changes in milk price paid by fluid milk processors and changes in mark-ups. In this section, we present the analysis results regarding the impacts of these two factors.

Impact of the Compact on the Retail Milk Price

We first examined the changes in retail milk prices in Boston and Hartford using the ARMA model presented in the previous section. A model specification test suggests that there is no integration of the process and therefore we use an ARMA model rather than an autoregressive integrated moving average (ARIMA) model. Using the Akaike Information Criteria (AIC), we choose the orders, p and q, of the autoregressive and moving-average process. Table 2 presents the maximum likelihood estimates of the model, ARMA, for the retail milk prices in Boston and Hartford.

Both the autoregressive and moving-average orders of the process include 1-month and 12-month lags, which implies that the retail milk price is affected by its levels as well as the unexplained disturbances in the last month and in the same month last year. We also include the Class I milk price as an additional explanatory variable. It is specified in the model with its values in the current month, the last month, and the same month last year. The estimated models fit the data very well.

Estimates of the first-order autoregressive parameter are 0.937 for Boston and 0.901 for Hartford, and both are significant at the 0.95 level. This implies that retail milk price in any particular month is significantly and positively affected by its level in the previous month. The estimate of the 12^th-order autoregressive parameter is negative for both cities and significant only for Hartford. It means that its level in the last year plays a role of dampening the large effect of the one-month previous level on the current level. The estimated parameter for the Class I milk price at the current month is positive for both cities but significant for Boston only. The parameter of Class I milk price in the last month is not significant and the parameter of Class I price at the same month in previous year is significant for both cities.

Based on the estimation results in Table 2, we obtain forecasts of retail prices (P^j*) over the post-Compact period under the assumption of no Compact and no exogenous shock in retail prices. The results of forecasts and confidence intervals for test statistics are shown in Table 3 for Boston and Table 4 for Hartford. The actual net price paid by WIC is the retail price less the compact over-order premium. The difference between the forecasted retail price and the actual net price paid by the WIC programs indicates how the actual net price paid by WIC programs differed from what would have been expected during July 1997 to June 1998. If the difference is positive and statistically significant, it implies that the net price paid by WIC programs is lower than what would be expected based on historical relationships between the retail price and the Class I milk price. On the other hand, if the difference is negative and statistically significant, it implies that the net price paid by WIC programs is higher than what would be expected based on the historical relationships. A difference that is not statistically significant implies that we can not determine the difference although it may exist. The negative and statistically significant differences are of greater interest because higher net prices to WIC programs may adversely affect milk availability to program participants.

For Boston, the actual net price paid by the WIC program is not statistically significantly different than the price predicted by the model in the absence of the Compact (Figure 2). This suggests that the net price paid by the WIC program in Massachusetts—within the limits of statistical uncertainty—was not affected by the price regulation under the Compact. For Hartford, the net price paid by the WIC program was significantly lower than the predicted value in the first month of the Compact but significantly higher than the predicted value during the remaining months except June 1998 (Figure 3). This provides evidence that the net price paid by the WIC program in Hartford during much of the first year of compact was higher than would have been expected based on historical relationships.

Changes in Mark-up Behavior

As discussed in the previous section, changes in the retail milk prices in the post-Compact period are likely due to at least two major factors: the compact over-order premium and changes in mark-up behavior. We now examine whether there has been a change in mark-up behavior by wholesalers and processors in the Boston and Hartford markets. Then we compare the observed mark-up with the mark-up expected in the absence of Compact price regulation to assess impacts of mark-up on the net price paid by WIC programs. These two analyses are related: a change in mark-up behavior helps to explain the underlying reasons for changes in the net price paid by WIC.

Similar to the impact analysis of the over-order premium on retail prices, we examine the changes in the mark-up using an ARMA (1,1) model. Table 5 presents the estimated model for Boston and Hartford. For both cities, estimates of autoregressive parameters are significant. The estimates of moving average parameter are significant for Boston but not significant for Hartford. To explain the mark-up process, we add some explanatory variables: the level of the Class I price in the current month, previous month, and the same month of last year, and a dummy variable (D₉₇). D₉₇ is 1 for months since January 1997 and 0 otherwise. The dummy variable is to capture the jump in mark-up over the period. Estimates of the dummy variable are significant and positive for both cities as expected. The current level of Class I price has a significant and negative effect on the mark-up. This is consistent with the observations that the mark-up of wholesalers and retailers is generally negatively related to Class I milk price.

One way to examine whether there has been a change in mark-up behavior is to compare the observed mark-up with the mark-up that would have been expected given a change in the processors’ fluid milk cost. To do this, we compare the observed mark-up with the mark-up that our model predicts given the increased cost of fluid milk to processors under the Compact price regulation. Based on historical relationships, the model predicts an increase in the cost of fluid milk to processors will result in a decrease in the mark-up, at least in the short run. Thus, based on historical price relationships, the increase in prices paid by processors due to the Compact over-order premium is predicted to result in a decrease in mark-up.

The decrease in predicted mark-up when price regulation under the Compact became effective in July 1997 is shown for Boston and Hartford in Figure 4 and Figure 5. However, the actual mark-up remained about the same as in previous months, despite the increase in the cost of milk to processors. The difference between the predicted mark-up and the observed mark-up was statistically significant for Boston for the first two months of Compact but significant for Hartford for the whole period (July 1997 to June 1998). This provides evidence that retailers and wholesalers in both markets engaged in mark-up behavior at times during the Compact different from their mark-up behavior in the years before the Compact. Retailers and wholesalers maintained their mark-up for fluid milk by raising milk prices when price regulation under the Compact took effect, rather than let margins fall as they appear to have done during 1990 to mid-1997.

Given the evidence of changes in mark-up behavior by wholesalers and processors, we wish to examine whether the actual mark-up differed from the mark-up that would have been expected in the absence of Compact price regulation. The difference between these two quantities provides additional evidence about the changes in the net price paid by WIC, as indicated by equation (5). To make this comparison, we calculate the difference between the actual mark-up with the mark-up predicted with our model using the Class I price (i.e., without the Compact over-order premium). Consistent with the results for the analysis of the net price paid, there is no statistically significant difference between the actual and predicted values of the mark-up in the Boston market (Figure 6). Thus, there is little evidence from our analysis of mark-up that the net price paid by the WIC program in Massachusetts increased after the onset of Compact price regulation.

In Hartford, however, the story is different. During the first three months of Compact price regulation, the predicted mark-up was statistically significantly higher than the actual mark-up observed in the Hartford market (Figure 7). This provides evidence that, during those months, the net price paid by WIC was lower than it would have been in absence of the Compact. From November 1997 to May 1998, however, the actual mark-up in Hartford was statistically significantly higher than the predicted values. This provides evidence that the net price paid by the WIC program was higher than it would have been in the absence of the Compact. The analyses summarized in Figures 3 and 7 together indicate that the higher net price paid by the WIC program in Connecticut resulted from higher than predicted mark-ups in the Hartford market.

Conclusions

Price regulation under the Compact has been in effect since July 1997. Although its main objective is to improve the sustainability of dairy farms, it has been associated with increases in the retail milk price, particularly during the first month of price regulation. This study examines the causes of the increases in retail milk prices in Boston and Hartford and the potential effects on the WIC programs. In general, the total number of WIC participants in the compact states decreased slightly in the first three months of the Compact but increased over the period of October 1997 to February 1998. This suggests that the number of participants in WIC programs has not been significantly affected by the Compact.

Our statistical models suggest that the retail milk price and mark-up behavior changed significantly in the first two months of the Compact in Boston and for most of the months in Hartford. In the Boston market, the analyses of retail prices and mark-ups indicate that there was no statistically significant difference between the actual net price paid by WIC and the net price predicted in the absence of the Compact. This provides evidence that reimbursement of the Compact over-order premium by the Compact Commission is helping to avoid an increase in the cost of milk to Massachusetts WIC programs. On the other hand, in the Hartford market, both the retail price and mark-up models indicate that the net prices paid by the Connecticut WIC program were greater than the predicted values from November 1997 to May 1998. Thus, despite reimbursement of the Compact over-order premium, it appears that Connecticut WIC program participants in the Hartford area paid higher net prices for milk than they would have in the absence of the Compact.

These higher net prices resulted primarily from a change in mark-up behavior by wholesalers and retailers in the Hartford market in response to Compact price regulation. A possible explanation for the difference between Boston and Hartford is the differences in market concentration and competition. In order words, the Boston market may be more competitive or more efficient and therefore both the retail milk price and mark-up are relatively lower than that in Hartford. Future studies can benefit more explicit treatment of these factors.

References

Bailey, Ken, and Jose Gamboa. A Regional Economic Analysis of Dairy Compacts: Implications for Missouri Dairy Producers. Commercial Agricultural Program, University of Missouri, 1999.

Greene, W. Econometric Analysis, New York: NY: Macmillan Publishing Company.

Knutson, Ron. “Compacts Create Winners and Losers.” Hoard’s Dairyman. April 10, 1999.

Pindyck, Robert S. and Daniel L. Rubinfeld. Econometric Models and Economic Forecasts. Third ed., New York: McGraw-Hill Inc., 1991.

USDA, Agricultural Marketing Service, Dairy Programs, Market Information. Unpublished Information. Washington, D. C., 1998.

Table 1. The Number of WIC Participants in the Compact States from June 1997 to February 1998

Year	Month	States						All Compact States
		CT	ME	MA	NH	RI	VT
1997	June	57756	26737	114573	19365	22967	16168	257566
	July	59013	26572	113663	19079	22739	16232	257298
	August	60752	26423	111906	19140	22567	16197	256985
	September	60738	26359	111311	18952	22628	16411	256399
	October	62550	25395	121555	18960	22867	16452	267779
	November	62680	26382	121300	18737	21725	16329	267153
	December	62253	26152	120907	18632	21951	16361	266256
1998	January	62986	26119	122505	18817	22697	16337	269461
	February	59210	25786	120915	18605	22712	16345	263573

Table 2. Estimation Results of the ARMA Model of Retail Milk Prices for Boston and Hartford (January 1990 to June 1997)

	Boston			Hartford
Process	Parameter		Standard error	Parameter		Standard error
Intercept	1.829	**	0.149	2.118	**	0.087
Autoregressive Parameters
zt-1	0.937	**	0.055	0.901	**	0.065
zt-12	-0.051		0.059	-0.112	**	0.060
Moving Average Parameters
et-1	0.307	**	0.134	0.395	**	0.145
et-12	0.095		0.148	-0.066		0.154
Class I milk price^1)
Pt^I	0.261	**	0.098	0.077		0.058
Pt-1^I	0.043		0.100	0.083		0.058
Pt-12^I	0.174	**	0.084	0.090	**	0.047
Estimate of Variance	0.0017			0.0004
AIC	-311.23			-371.86.41

¹⁾ P_t^I, P_t-1^I, and P_t-12^I are the Class I milk prices at the current month, at the last month, and at the same month in

previous year.

**P<.05.

Table 3. Impacts on the net price paid by the WIC program in Boston¹⁾

Year	Month	Actual			Forecasted retail price^4)	Difference^5)	Confidence interval^6)
		OOP^2)	Retail price	Net price paid by WIC^3)			Lower	Upper
1997	January		2.42	2.42	2.37
	February		2.45	2.45	2.39
	March		2.45	2.45	2.44
	April		2.45	2.45	2.46
	May		2.45	2.45	2.46
	June		2.44	2.44	2.44
	July	0.259	2.64	2.38	2.43	+0.05	2.36	2.50
	August	0.255	2.63	2.37	2.44	+0.07	2.36	2.53
	September	0.245	2.63	2.39	2.46	+0.07	2.37	2.55
	October	0.141	2.62	2.48	2.50	+0.02	2.40	2.60
	November	0.078	2.63	2.55	2.53	-0.02	2.42	2.63
	December	0.073	2.63	2.56	2.52	-0.04	2.41	2.63
1998	January	0.064	2.60	2.54	2.48	-0.06	2.36	2.59
	February	0.035	2.59	2.55	2.47	-0.08	2.35	2.59
	March	0.039	2.60	2.56	2.48	-0.08	2.36	2.60
	April	0.033	2.60	2.57	2.49	-0.08	2.36	2.62
	May	0.077	2.60	2.52	2.48	-0.04	2.35	2.61
	June	0.146	2.54	2.39	2.44	+0.05	2.31	2.57

¹⁾ Based on estimates of the ARMA model over January 1990 to June 1997.

²⁾ The Compact over-order premium per gallon, which is calculated by dividing the over-order premium per cwt by 11.6.

³⁾ Equals the payment for fluid milk by WIC programs after the Compact Commission’s reimbursement for over-order premium.

⁴⁾ Predicted values for the pre-compact period.

⁵⁾ The difference between forecasted retail price and the net price paid by WIC. Positive numbers indicate that the predicted prices without the

compact are greater than the actual net prices paid by the WIC programs and the negative numbers indicate that the predicted prices without the

compact are less than the actual net price paid by the WIC programs.

⁶⁾ Represents the range of values expected for predictions of the net price paid by WIC based on the ARMA model.

Table 4. Impacts on the net price paid by the WIC program in Hartford¹⁾

Year	Month	Actual			Forecasted retail price^4)	Difference^5)		Confidence interval^6)
		OOP^2)	Retail price	Net price paid by WIC ^3)				Lower	Upper
1997	January		2.51	2.51	2.48
	February		2.49	2.49	2.48
	March		2.49	2.49	2.48
	April		2.49	2.49	2.49
	May		2.49	2.49	2.49
	June		2.49	2.49	2.49
	July	0.259	2.68	2.42	2.48	+0.06	*	2.44	2.52
	August	0.255	2.68	2.42	2.47	+0.05		2.42	2.52
	September	0.245	2.68	2.44	2.47	+0.03		2.42	2.52
	October	0.141	2.68	2.54	2.48	-0.06	*	2.43	2.53
	November	0.078	2.68	2.60	2.49	-0.11	*	2.44	2.55
	December	0.073	2.68	2.61	2.48	-0.13	*	2.43	2.54
1998	January	0.064	2.68	2.62	2.46	-0.16	*	2.40	2.52
	February	0.035	2.68	2.64	2.45	-0.19	*	2.39	2.51
	March	0.039	2.68	2.64	2.45	-0.19	*	2.39	2.51
	April	0.033	2.68	2.65	2.45	-0.20	*	2.39	2.51
	May	0.077	2.68	2.60	2.44	-0.16	*	2.38	2.50
	June	0.146	2.61	2.46	2.42	-0.04		2.36	2.48

¹⁾ Based on estimates of the ARMA model over January 1990 to June 1997.

²⁾ The Compact over-order premium per gallon, which is calculated by dividing the over-order premium per cwt by 11.6.

³⁾ Equals the payment for fluid milk by WIC programs after the Compact Commission’s reimbursement for over-order premium.

⁴⁾ Predicted values for the pre-compact period.

⁵⁾ The difference between forecasted retail price and the net price paid by WIC. Positive numbers indicate that the predicted prices without the

compact are greater than the actual net prices paid by the WIC programs and the negative numbers indicate that the predicted prices without the

compact are less than the actual net price paid by the WIC programs.

⁶⁾ Represents the range of values expected for predictions of the net price paid by WIC based on the ARMA model.

Table 5. Estimation Results of the ARMA Model of Mark-up in Boston and Hartford (January 1990 to June 1997) ²⁾

	Boston Model			Hartford
Process (zt)	Parameter		Standard error	Parameter		Standard error
Intercept	1.656	**	0.161	2.035	**	0.093
Autoregressive Parameters
zt-1	0.945	**	0.046	0.827	**	0.093
Moving Average Parameters
et-1	0.338	**	0.127	0.249		0.249
Class I Price^1)
Pt^I	-0.630	**	0.101	-0.861	**	0.062
Pt-1^I	0.078		0.093	0.081		0.059
Pt-12^I	0.153	*	0.076	0.092		0.049
D97	0.095	**	0.038	0.051	**	0.019
Estimate of Variance			0.0012			0.0004
AIC			-296.25			-377.02

¹⁾ P_t^I, P_t-1^I, and P_t-12^I are the Class I milk prices at the current month, at the last month, and at the same month in previous year. D₉₇ is a dummy variable, which

equals one for all the months since January 1997 and zero for other periods.

²⁾ The mark-up is defined by the difference between retail prices and the Class I milk price.

*P<.10., **P<.05.

Forecast is the predicted value of the retail price without the Compact. U95 and L95 indicate upper and lower 95% confidence intervals for the prediction.

Forecast is the predicted value of the retail price without the Compact. U95 and L95 indicate upper and lower 95% confidence intervals for the prediction.

Forecast is the predicted value of the mark-up with the Compact. U95 and L95 indicate upper and lower 95% confidence intervals for the prediction.

Forecast is the predicted value of the mark-up with the Compact. U95 and L95 indicate upper and lower 95% confidence intervals for the prediction.

Forecast is the predicted value of the mark-up without the Compact. U95 and L95 indicate upper and lower 95% confidence intervals for the prediction.

Forecast is the predicted value of the mark-up without the Compact. U95 and L95 indicate upper and lower 95% confidence intervals for the prediction.