SYMPOSIUM: ECONOMIC AND MARKETING FACTORS IN THE FLUID MILK INDUSTRY Factors Affecting Labor Productivity and Cost

per Gallon in Fluid Milk Plants

E. M. ERBA and R. D. APLIN
Department of Agricultural, Resource and Managerial Economics, Cornell University, Ithaca, NY 14853

ABSTRACT

The fluid milk industry represents the largest sector of the US dairy industry and processed more than 5 billion gal (1 gal = 3.88 L) of milk in 1994. The industry tends to be more motivated by cost competition than by product or market development, and, hence, ongoing interest exists for studies of processing and distribution costs. Furthermore, recent developments in technology and processing equipment, which extend product shelf-life and permit far-reaching distribution systems, have implications for the traditional market orientation and structure of the industry. Results from studies of processing costs for fluid milk have a variety of uses. Management and executive personnel of fluid milk plants may apply the results t o their own operations t o gauge the performance of their operations against other similar milk plants. Such studies may also reveal which aspects of fluid milk operations offer the most benefit from internal restructuring or capital investments. The results may also be useful for purposes of state control regulation of milk, especially for states that regulate milk prices at the wholesale or retail levels. Studies on the cost of processing have also been invaluable in providing cost components t o model the dairy industry and to predict changes in structure. During the past 35 yr, the cost of processing fluid Abbreviation key: CAP = sustainable plant capac- milk has been analyzed several times. Studies by ity, GEQ-M = gallon equivalents processed per Blanchard et al. ( 2 ) and Bond (3) partitioned plants month, GHG = percentage of processing volume pack- into separate cost centers and used cost data to anaaged in gallon and half-gallon containers, OLS = lyze differences in efficiencies among participating ordinary least squares, PCAP = percentage of plant plants. Other research (8, 9, 12) has investigated capacity utilized, TCL = degree of automation and processor sales, costs of goods, operating costs, and technology in the cooler and load out area, TPF = gross and net margins for fluid milk plants of moderate size. Because of difficulties encountered in recruiting participants in processing cost studies or from a lack of adequate numbers of representative plants, Received (June 27, 1995. Accepted December 27, 1995 economic engineering studies have served as an alter1996 J Dairy Sci 79:1304-1312

Regression analyses were performed on data from 13 fluid milk plants to identify the factors that affect plant labor productivity and plant cost per gallon. Explanatory variables included basic plant descriptors, such as type of plant ownership, gallons processed per month, level of automation and technology in the processing and filling area, plant capacity utilization, and labor cost per hour. The results suggested that higher labor cost per hour, more technologically advanced equipment in the processing and filling area, and a higher percentage of product packaged in gallon and half-gallon containers have positive effects on the productivity of plant labor. Vertically integrated plants that were owned by supermarket chains were predicted to have higher rates of productivity per hour of labor than proprietary or cooperative plants. Type of plant ownership appeared to have the most significant effect on plant cost per gallon. Vertically integrated plants owned by supermarket chains were predicted to realize lower costs per gallon than cooperative or proprietary plants. Cooperative plants were predicted to have higher costs per gallon than proprietary plants. Despite the positive effect of labor cost per hour on labor productivity, higher cost of labor per lhour was predicted t o increase plant cost per gallon ( Key words: fluid milk, plant capacity, labor productivity, processing cost)

degree of automation and technology in the processing and filling area.

INTRODUCTION

1304

SYMPOSIUM: ECONOMIC AND MARKETING FACTORS IN THE FLUID MILK INDUSTRY

1305

TABLE 1. Description ofvariables and form of variables used in regression analyses of factors affecting fluid milk plant labor productivity and plant cost per gallon. Variable Captive Cooperative GEQ-M Maximum CAP Wage PCAP GHG Production cost TCL TPF Form Indicator Indicator Polynomial Polynomial Linear Linear Linear Linear Linear Reciprocal Reciprocal Description Vertically integrated plant owned by supermarket chain Plant owned by milk marketing cooperative Gallon equivalents processed per month Maximum sustainable plant capacity in gallon equivalents per month Average cost of wages and benefits for plant employees Percentage of plant capacity utilized Percentage of processing volume packaged in gallon and half-gallon containers Gallon equivalents of product processed per hour of plant labor Plant cost per gallon in cents per gallon equivalent Degree of automation and technology in cooler and load out area Degree of automation and technology in processing and filling area

native method of estimating the minimum processing sistent with this plant profile agreed to participate in costs achievable per gallon and investigating the con- the study: 2 vertically integrated plants owned by sequences of various plant sizes on processing costs supermarket chains (captive plants), 2 cooperatively ( 4 , 5 , 7, 11, 13). owned plants, and 9 proprietary dairies. None of Studies that attempt to identify the factors that those plants produced only fluid milk products. Other affect plant productivity and the cost of processing are beverage products included coffee creamers; half and less ubiquitous. Thraen et al. ( 1 4 1 estimated a func- half; heavy, medium, and light creams; ice cream mix; tional relationship between total plant cost and plant juices; buttermilk; flavored drinks; and bottled water. volume based on data from 15 cooperatively owned The data collected relied on monthly observations and operated fluid milk plants, suggesting that per over a 12-mo period during 1993 and 1994 (Table 1). unit costs decrease as plant processing volume in- Basic plant descriptors, such as gallon equivalents creases. Aplin ( 1) indicated that economies of scale, processed per month ( GEQ-M) , maximum sustainautilization of plant processing capacity, product mix, ble plant capacity in gallon equivalents per month and level of technology in the processing and cooler (maximum C A P ) , and mean cost of wages and areas would be expected to influence the cost of benefits for plant employees were collected directly processing as well as plant productivity per hour of from the plants. Other measures, such as percentage labor. of plant capacity utilized ( PCAP), percentage of This study identifies the factors that cause varia- processing volume packaged in gallon and half-gallon tion in labor productivity and plant cost per gallon in containers ( GHG), gallon equivalents of product fluid milk plants. The results of the study are intended to be used by managers of fluid milk plants processed per hour of plant labor (production), and that are similar in size and structure to the milk plant cost per gallon in cents per gallon equivalent (cost), were calculated from submitted information. plants participating in the study. The degree of automation and technology in the cooler and load out area ( TCL) and the degree of automaMATERIALS AND METHODS tion and technology in the processing and filling area ( T P F ) were evaluated by the plant manager a t each Source of Data participating plant. A 10-point scale was used to asA survey of fluid milk plants in the northeastern sess the level of technology (1 = the lowest level of and middle Atlantic regions of the US was conducted technology, and 10 = the latest, most innovative techto assess the cost of processing fluid milk products nology). Table 2 lists selected summary statistics for and to identify the factors that cause plant cost and the dependent and independent variables, based on productivity to vary. The study targeted fluid milk the data collected from the 13 fluid milk plants. The plants wiith monthly processing volumes of a t least mean, standard error, and minimum and maximum 1.5 million gal/mo, effective management styles, high values are given to provide a numerical description of labor productivity, significant market presence, and each variable. Although only fluid products processed by the innovative or technologically advanced processing and cooler areas. Thirteen plants with characteristics con- plants were used to determine the volumes processed
Journal of Dairy Science Vol. 79, No. 7,1996

1306

ERBA AND APLIN

each month, nearly all plants purchased finished products to be distributed to their customers. Some of the finished products were juices or flavored drinks, but a significant portion of the products were solid foods, such as yogurt, sour cream, cottage cheese, and cream cheese. Clearly, these items were not measurable in gallon equivalents. Moreover, productivity and cost measures were more sensible if based on gallons per hour or cost per gallon. Although using GEQ-M excluded purchased finished products, only about 7% (range: 0 to 12%) of the total number of cases handled in the cooler per month were purchased from other food manufacturers or distributors. Consequently, the effect of omitting purchased products from the analyses was expected to be minimal.
Model Specification

Although the analysis afforded the flexibility of a variety of model specifications and functional forms, the two models developed were multiple regression models. The equation used to model production was specified as

and the equation used to model cost was specified as costit =

820 + P21wageit + Pz2ca~tivei

+ +

P23coopi + @24maximumCAPit

/325maximumCAP?t + P26maximumCAP~t

,!3271n(productionit) + eit

121

the volume of product processed per month, measured in gallon equivalents, divided by the total monthly hours logged by processing, filling, cooler, and load out labor, as well as any other plant labor that was not assigned to a specific cost center (for example, maintenance, quality control, and plant management). The hours worked by plant employees assigned to the blow-molding area, sales representatives, and general and administrative personnel were not included in production. The dependent variable, cost, accounted for the cost of processing, filling, cooler, load out labor, and any other plant labor that was not assigned to a specific cost center; depreciation costs from all plant equipment and structures; utilities; plant maintenance and repairs; cleaners and lubricants; plant supplies; pest control; refuse collection; security; leases; property taxes; and insurance. The cost of labor for blow-molding, packaging materials, ingredients, distribution, selling expenses, and general and administrative expenses were not included in cost. Some of the variables were hypothesized to impart nonlinear effects on the dependent variables. For example, the effect of TPF on production was not likely to be constant over the range of the reported data. The explanatory variables TPF and TCL were hypothesized to have decreasing returns to improvements. For example, using a 10-point scale, improving TPF from 2 to 4 was expected to have a greater impact on production than improving TPF from 7 to 9. Thus, TPF and TCL were specified as reciprocal variables. Plant size, as measured by GEQ-M and maximum CAP, was also expected to impart nonlinear effects on the dependent variables such that larger volumes processed increased production and decreased cost. Third-degree polynomials were specified for GEQ-M and maximum CAP to allow more flexibility when the effects of these two explanatory variables were estimated. To capture the nonlinear effect of production on cost, which was hypothesized to decrease a t a decreasing rate, a logarithmic term was specified for production in Model 121.

where captive = captive milk plants, coop = cooperative milk plants, i = 1,. . . ,13 plants, t = 1, . . . ,21 Recursive System time periods, and uit and eit were random disturThe conventional wisdom concerning the relationbances with E(uit, ujt) = 0 for i f j; E(qt, Ejt) = 0 f o r i # ship of cost and production might suggest that a j; E(uit, fit) = 0 for all i; uit = piuit-1 + +it; f i t = 6ieit-1 + simultaneous equation problem was encountered with oit; uit - N(0, and f i t - N(0, I&. Models 1 1 1 and [21. Because cost was likely to be The dependent variables, production and cost, were influenced by production but production was not calculated from survey data to ensure that compara- likely to be influenced by cost, Models [ l l and [21 were ble cost and labor productivity figures were analyzed. cast in the framework of a recursive model. In a The dependent variable, production, was defined as recursive system, applying ordinary least squares

&;

Journal of Dairy Science Vol. 79,No. 7,1996

SYMPOSIUM: ECONOMIC AND MARKETING FACTORS IN THE FLUID MILK INDUSTRY

1307

TABLE 2. Summary statistics for variables used in regression analyses of factors affecting fluid milk plant labor productivity and plant cost per gallon. Variable' GEQ-M ( ~ 1 0 6 ) Maximum CAP ( ~ 1 0 ~ ) Wage CAP GHG Production cost TCLz TPF2 Mean

SE
1.4

Minimum
1.8 2.6

Maximum
9.6 9.7 28.2 100.0 96.1 305.2 29.8 10.0

3.6
4.8 21.2

75.4
80.6 159.1 22.6
5.1

2.0 4.5 11.0

9.5
52.0 54.6
92.7

9.9
50.0

3.9 3.0
1.4

13.3
1.0 4.0

6.8

9.0

1GEQ-M = Gallon equivalents processed per month, maximum CAP = maximum sustainable plant capacity in gallon equivalents per month, wage = average cost of wages and benefits for plant employees, CAP = percentage of plant capacity utilized, GHG = percentage of processing volume packaged in gallon and half-gallon containers, production = gallon equivalents of product processed per hour of plant labor, cost = plant cost per gallon in cents per gallon equivalent, TCL = degree of automation and technology in the cooler and load out area, and TPF = degree of automation and technology in the processing and filling area. 2The TCL and TPF were evaluated by the plant manager a t each participating plant. A 10-point scale was used to assess the level of technology ( 1 = the lowest level of technology to 10 = the latest, most innovative technology).

t o each equation separately leads to unbiased and consistent estimates ( 6 ) . The right-hand side of Model [l]contained only exogenous variables, and because those variables were assumed to be uncorrelated with the disturbance term, uit, OLS was applied directly. Model [2] contained the endogenous variable production as an explanatory variable along with the nonstochastic explanatory variables. Because vit was assumed t o be uncorrelated with fit, production was a predetermined variable insofar as cost was concerned, and Model [21 was estimated using OLS.

( OLS)

and, from the residuals, individual plant variances were calculated. The transformation was completed by by dividing both sides of Model [ll and Model [2J the individual plant standard deviations. The transformed equations were estimated by OLS, utilizing all of the pooled observations.
RESULTS AND DISCUSSION Plant Labor Productivity

The 13 fluid milk plants processed and packaged a mean of 159 g a m (range: about 93 to 305 g a l h ) . The regression results presented in Table 3 supported Pooling of Cross-Sectional and Time-Series Data conventional wisdom regarding production-plants that hire labor at a higher wage, plants with higher To generate a sufficient number of observations t o GHG, plants with higher TPF, and captive plants perform the regression analyses, cross-sectional and were expected to realize gains in production. time-series data were pooled. Under such an apThe specified statistical model appeared to fit the proach, regression disturbances were assumed to be data well. The R2 for the model was 86.3%,and all of mutually independent but heteroskedastic and au- the signs on the coefficients coincided with prior extocorrelated. First-order coefficients for autocorrela- pectations. Two variables, PCAP and TCL, were not tion, p and 6, were estimated for each plant by statistically significant at a reasonable level. regressing uit on Uit-1 and it on fit-1 ( 6 ) . The results Although the mean wage was $21.20/h, Table 2 revealed no statistical evidence to warrant data indicates substantial differences in wages among the transformations to correct for first-order autocorrela- participating plants. Plants that hired labor at a tion in Models [ l l or 121. higher wage were predicted to experience increases in Many possible remedial measures were available to production. For example, a 5.3 g a l h increase in correct for heteroskedasticity, and the transformation production was predicted or plants that hired labor used resulted in regression coefficients that were con- at $l.OO/h more than the mean labor cost. sistent and unbiased with minimum variance asympThe results supported the contention that higher totically ( 10). First, OLS was applied to all observa- GHG increases production. The inverse implication tions for each regression model. The OLS estimates was that packaging products in smaller container were then used to calculate the regression residuals, sizes reduced production. Filling machinery for quart
Journal of Dairy Science Vol. 79, No. 7, 1996

1308

ERBA AND APLlN

TABLE 3. Regression results for plant labor productivity. Variable2 Constant Wage GHG Captive Cooperative TPF TCL GEQ-M (GEQ-MI2 (GEQ-MI3 PCAP Form Linear Linear Indicator Indicator Reciprocal Reciprocal Linear Squared Cubed Linear Coefficient
-0.97 5.26 0.35 102.61 -26.69 -145.32 -6.83 39.42 -9.42 0.61 -0.03

SE
0.66 0.32 0.17 6.99 5.08 41.05 -1.22 14.87 3.10 0.19 0.15

<o. 0001
0.0380 <0.0001 <0.0001 0.0005 0.2247 0.0089 0.0029 0.0015 0.8496

0.1417

1The regression was based on 156 observations, and the R2 for the model was 0.863. W a g e = Average cost of wages and benefits for plant employees, GHG = percentage of processing volume packaged in gallon and half-gallon containers, captive = advantage or disadvantage of captive versus proprietary plant, cooperative = advantage or disadvantage of cooperative versus proprietary plant, TPF = degree of automation and technology in the processing and filling area, TCL = degree of automation and technology in the cooler and load out area, and GEQ-M = gallon equivalents processed per month, and PCAP = percentage of plant capacity utilized.

(0.946 L), pint (0.474)) and half-pint (0.237 L) containers operated at speeds of 100 t o 120 units/min; the speed of filling machinery for half-gallon and gallon containers ranged from about 80 to 100 units/ min and ,about 60 to 75 unitdmin, respectively. Although half-gallon and gallon fillers produced fewer unitdmin, the difference in total volume of product packaged rnore than compensated for the slower filling speed. Despite similarities in plant characteristics for the participating companies, substantial variation in GHG was evident. The range in GHG was 54.6% t o 96.1%; the mean was 80.6% (Table 2 ) . Although the estimated coefficient for GHG was statistically significant a t a reasonable level, the practical significance of GHG on production was not exceptional. The regression results suggested that a 10% increase in GHG increased production by 3.5 g a m . Type of plant ownership also appeared t o have significant ramifications for production. An advantage of 102.6 g a m was predicted for captive plants compared with that of full-line proprietary milk plants (Table 3). A disadvantage of 26.7 g a m was predicted for cooperative plants compared with proprietary milk plants, and cooperative plants were predicted t o process 129.3 g a m less than captive plants. The tremendous advantage in production predicted for captive plants might not be surprising, but, to those who are not familiar with the industry, the reasons for the advantage might not be readily apparent. Narrower product mixes are maintained by captive plants, which process fewer products under fewer labels and use fewer packaging sizes. Specifically, most products are packaged in gallon and half-gallon containers:, and only a small percentage of products
Journal of Dairy Science Vol. 79, No. 7, 1996

are packaged in quart, pint, or half-pint containers. Some of the advantage realized by captive plants in this respect was captured by the explanatory variable GHG. Relative to the total number of products handled, few (if any) finished products from outside sources are brought into the coolers of captive plants for distribution, reducing the number of products in the cooler and simplifylng filling of orders and load out procedures. For distribution, captive plants serve supermarket stores that place orders for similar mixes of products with little variation in order size. In combination, the characteristics described point toward operations with high product turnover and high labor productivity, which are inherently easier to manage. Similar to proprietary fluid milk plants, cooperative plants tend t o operate as full-line processing facilities. The analysis of the 13 fluid milk plants suggested that cooperative plants had lower production, but the structure of cooperatively owned and operated businesses was not likely to be a valid reason for the result. Some of the top food processing businesses in the US are cooperatives. Consequently, other reasons must be sought to explain why cooperative plants were predicted t o experience lower production. First, because the primary owners of a dairy cooperative are dairy farmers, raising equity capital to invest in the fluid milk facilities can be difficult. Second, dairy cooperatives are primarily responsible for marketing the milk of their members, and some of the milk produced might be sold to other milk processing companies. The remaining supply might need to be apportioned among other plants owned by the cooperative. Under those conditions, a cooperative

SYMPOSIUM: ECONOMIC AND MARKETING FACTORS IN THE FLUID MILK INDUSTRY

1309

10

TPF
Figure 1 Relative effect of degree of automation and technology in the processing and filling area ( T P F ) on gallon equivalents of product processed per hour of plant labor. The vertical axis was adjusted such that the minimum predicted effect of TPF was zero gallons per hour. Plant managers a t each participating plant evaluated TPF. A 10-point scale was used to assess the level of technology (1 = the lowest level of technology to 10 = the latest, most innovative t.echnology).

but increasing TPF from 6 t o 9 was predicted to increase production by 8 g a m . By industry standards, the participating plants were large. The plants processed and packaged an average of 3.6 million gal/mo with a range of 1.7 million gaVmo to 9.6 million gal/mo. Although the size of the plants in the study was not representative of fluid milk plants throughout the US, plant size, as measured by GEQ-M, was investigated as a potential factor affecting production. The results revealed that GEQ-M had a statistically significant nonlinear relationship with production, but the relationship did not follow a typical economies of scale pattern (Figure 2 ) . As GEQ-M increased from about 1.7 million to 3 million, production increased by about 6 gal/h. However, increasing GEQ-M from 3 million to 7 million decreased production by about 27 g a m . Two variables, PCAP and TCL, were not statistically significant at a reasonable level, and evidence was insufficient to support the contention that plants that utilize their capacity more fully or plants that have more modern cooler and load out areas experienced higher production.
Plant Cost per Gallon

plant might be forced to operate at below optimal capacity, despite the negative effects on production and cost. Finally, cooperative plants typically balance the milk supply in a given market for the economic benefit of their members. Performance of this balancing function results in additional costs, and perhaps less efficient use of labor and facilities. The eflkcts of the nonlinear variables on production were not obvious when the coefficients listed in Table 3 were reviewed. Graphs for each of the statistically significant nonlinear variables were generated to provide a visual description of the effect on production. The vertical axis of each graph was scaled such that the minimum predicted effect of the nonlinear variable over the range of the data was zero. As such, the graphs provide a representation of the relative effect of the nonlinear variable on production. Positive impacts on production were predicted when TPF was increased in reasonable increments (Figure 1).The vertical axis measures the expected increase in production in gallons per hour, relative to the lowest reported level of TPF for the 13 plants. The hypothesized relationship for the two variables implied that the returns to production were largest at the lower range of TPF with h r t h e r increases in TPF resulting in much smaller improvements in production. The hypothesis was supported by the results. For examlple, increasing TPF from 4 to 7 was predicted to increase production by about 15 g a m ,

For the 13 fluid milk plants, cost averaged 22.6d gal and ranged from 13.3 to 28.2q/gal. The specified model for plant cost per gallon appeared to fit the data well. The R2 for the model was 98.1%, and all variables were significant at P = 0.05. The signs on captive and cooperative milk plants, wages, and

30

g 2 4

i-i
U

A
2
4

g 15

10

GEQ-M, million
Figure 2. Relative effect of gallons processed per month (GEQ-

M ) on gallon equivalents of product processed per hour of plant

labor. The vertical axis was adjusted such that the minimum predicted effect of GEQ-M was 0 g a h .
Journal of Dairy Science Vol. 79, No. 7,1996

1310

ERBA AND APLIN

TABLE 4. Regression results for plant cost per gallon.'

Variable2 Constant Wage Captive Cooperative Maximum CAP (Maximum CAPP (Maximum CAPP Production
~~ ~ ~~~~~

Form

Coefficient 1.06 0.93 -6.37 1.47

10.00

SE
0.32
0.06 0.79 0.50

P
0.0010 <0.0001
<0.0001 0.0038

...
Linear Indicator Indicator Linear Squared Cubed Logarithmic

-2.08 0.13 -2.24

~

1.77 0.32
0.02 0.80

<0.0001 <o ,0001

<0.0001 <0.0059

'The regression was based on 156 observations, and the R2 for the model was 0.981 'Wage = Average cost of wages and benefits for plant employees, captive = advantage or disadvantage of captive versus proprietary plant, cooperative = advantage or disadvantage of cooperative versus proprietary plant, maximum CAP = gallon equivalents processed per month, and production = gallon equivalents of product processed per hour of plant labor

production coincided with prior expectations. The remaining variable, maximum CAP, had a statistically significant nonlinear relationship with cost. The regression results presented in Table 4 partially supported the conventional wisdom regarding cost. Plants with higher production were predicted to have lower cost. Captive milk plants were predicted to have lower cost, but cooperative milk plants were predicted to have higher cost. Although the sign and magnitude of wage indicated an apparently unambiguous positive effect on cost, the total effect of wage on cost warranted further discussion. Differential effects on cost were expected as production varied, but the results showed that the relationship between cost and production was nearly linear (Figure 3 ) . Increasing production from 100 t o 150 gaVh was predicted to decrease cost by 0.9dga1,

and increasing production from 150 gal/h t o 200 g a m was predicted to decrease cost by 0.6q/gal. From the increases in producresults obtained from Model [l], tion were attainable by varying GHG, TPF, GEQ-M, and PCAP. Insofar as type of ownership was concerned, cost for captive milk plants was predicted to be lower than that for cooperative or proprietary plants, and cooperative plants were predicted to experience higher cost than proprietary dairies. The coefficients for captive and cooperative plants presented in Table 4 were net of the predicted effect of production on cost. Earlier results revealed that plant ownership affected production, and, in turn, production affected cost. Using the results from Models [ l l and 121, a total effect of plant ownership on cost was calculated. The advantage in production for captive milk plants equaled a decrease of 1.4@/gal in cost, and the results of Model [21 indicated that captive milk plants realized an additional 6.4dgal decrease in cost. Thus, cost was estimated to 3be 7.8q/gal lower for captive milk plants relative to proprietary plants. Similarly, the disadvantage in production for cooperative milk plants equaled an in cost, and Table 4 indicated that increase of 0.5@/gal cost was an additional 1.5dgal higher for cooperative milk plants. Thus, cost was estimated to be 2.0@/gal higher for cooperative plants relative t o proprietary plants and 9.8dgal higher than captive plants. Possible reasons for cost differences based only on type of ownership were identical to those offered to explain differences in production. One difference in specification for Models [ l l and __t_ [21 was the measure of plant size. Both maximum 0 50 100 150 200 250 300 350 CAP and GEQ-M served as proxies for plant size, but, Labor productivity, gal/h because depreciation costs for the plant structure and all of the equipment were included, maximum C A P Figure 3 . Relative effect of plant labor productivity on plant cost was deemed t o be a better representative Of plant size per gallon. 'phe vertical axis was adjusted such that the minimum for Model (21. predicted effect of production was Ou/gal.
Journal of Dairy Science Vol. 79,No. 7, 1996

SYMPOSIUM: ECONOMIC AND MARKETING FACTORS IN THE FLUlD MILK INDUSTRY

1311

31-

creasing TPF from 4 to 7 was predicted t o increase production by 15.5 g a m , resulting in a predicted decrease in cost by 0.2dgal.
CONCLUSIONS

4 6 Maximum CAP, million

10

Figure 4 Relative effect of maximum sustainable plant capacity in gallon equivalents per month (maximum CAP) on plant cost per gallon. The vertical axis was adjusted such that the minimum predicted el'fect of maximum CAP was Ou/gal.

The results show that the polynomial form of maximum CAP had a statistically significant nonlinear relationship with cost (Figure 4). As maximum CAP increased from about 2.2 million to about 4 million, cost was predicted to increase by about l.lq/gal. As maximum CAP increased from 4 million to 7.5 million, cost decreased by about 2.7q/gal. Although the initial section of the curve, which predicted increases in cost with increases in maximum CAP, was not expected, the latter section of the curve supported the contention that larger plants realize lower cost than do smaller plants. One variable, wages, was hypothesized t o impart separate and opposite effects on production and cost. This notion was supported by the regression results. For examlple, earlier results showed that a 5.3 gal/h increase in production was predicted for plants that hired labor at $l.OO/h more than the average labor cost. Because of the increase in production, cost was predicted to decrease by 0.08eIgal for each $l.OO/h increase in wages. However, the regression results for Model [2] indicated that the direct effect of hiring labor at $1.00 more than the average labor cost increased cost by 0.87q/gal. Thus, the total effect of wages on cost was an increase of 0.79q/gal for each $l.OO/h in.crease. Two variables, GHG and TPF, were predicted to have positive effects on production, but changes in either variable were predicted to have rather minor imp1icatio:ns for cost. For example, a 10% increase in GHG was predicted to increased production by 3.5 g a m , and increasing production by 3.5 gal/h was predicted 'to decrease cost by 0.05q/gal. Similarly, in-

In an industry that tends to be highly cost competitive, studies of plant productivity and processing costs are coveted by fluid milk plant management and executive personnel. The results may be used to gauge the performances of their operations against other similar milk plants and identify which aspects of their operations might benefit from internal restructuring or capital investments. This research goes beyond elementary comparisons of plant data to identify and to quantify the factors that explain variations in plant costs and labor productivity among plants. The results suggested that a variety of factors have significant implications for labor productivity. Type of plant ownership was identified as the most significant factor affecting labor productivity. However, higher labor cost per hour, more technologically advanced equipment in the processing and filling area, and a higher percentage of product packaged in gallon and half-gallon containers were also predicted to increase plant labor productivity. Type of plant ownership also appeared to have the largest effect on plant cost per gallon. Vertically integrated plants owned by supermarket chains were predicted to realize lower plant costs per gallon than proprietary plants. Cooperative plants were predicted to have higher plant costs per gallon than proprietary or vertically integrated plants. The total effect of higher cost of labor per hour was an increase in plant cost per gallon. Despite their positive effects on labor productivity, technologically advanced equipment in the processing and filling area and high percentages of product packaged in GHG containers were not characterized as factors that could reduce plant cost per gallon significantly.
REFERENCES
1Aplin, R. D. 1991. Factors contributing to profitability in fluid milk processing and distribution operations. Dairy Marketing Notes No. 1. Dep. Agric. Econ., Cornell Univ., Ithaca, NY. 2 Blachard, W. H., G. McBride, and A. L. Rippen. 1962. A cost analysis of fluid milk packaging operations. Tech. Bull. 285, Michigan State Univ. Agric. Exp. Stn., East Lansing. 3 Bond, G. C. 1978. Costs of processing and delivering milk in New Jersey, 1976-1977. S. R. 53, Dep. Agric. Econ. Marketing. New Jersey Exp. Stn., Cook Coll., Rutgers Univ., New Brunswick. 4 Criner, G. K., G. K White, and S. C. Howick. 1995. Fluid milk processing cost analysis. J . Dairy Sci. 78:1181. 5 Fischer, M., J . Hammond, and W. Hardie. 1979. Fluid milk processing and distribution costs. Bull. 530, Univ. Minnesota Agric. Exp. Stn., Minneapolis.
Journal of Dairy Science Vol. 79, No. 7, 1996

1312

ERBA AND APLIN

11 OConnell, P., and W. E. Snyder. 1964. Cost of analysis of fluid

6 Gujarati, I). N. 1988. Basic Econometrics. 2nd ed. McGraw-Hill Publ. Co., New York, NY. 7Jacobs, S. L., and G . K. Criner. 1990. Milk processing and distribution costs: the Maine model. Tech. Bull. 140, Univ. Maine Agric. Exp. Stn., Orono. 8 Jones, W., and F. Lasley. 1980. Milk processor sales, costs, and margins. USDAlEcon., Statistics, and Cooperatives Serv., 77. USDA, Washington, DC. 9 Jones, W. W. 1979. Milk processor-distributors sales, costs, and margins. Econ. Res. Serv., 66. USDA, Washington, DC. 10 Kmenta, J . 1971. Elements of Econometrics. MacMillan Publ. Co., Inc., New York, NY.

milk processing and distribution in Colorado. Tech. Bull. 86, Colorado State Univ. Exp. Stn., Fort Collins. 12 Pelsue, N. H., J r . 1992. Milk processing and distribution costs and returns. Fks. Rep. 65. Vermont Agric. Exp. Stn., Burlington. 13 Strain. J. R., and S. K. Christensen. 1960. Relationship between plant size and cost of processing fluid milk in Oregon. Tech. Bull. 55, Oregon State Coll. Agric. Exp. Stn , Corvallis 14Thraen, C. S., D. E Hahn, and J. B. Roof. 1987 Processing costs, labor efficiency, and economies of size in cooperativelyowned fluid milk plants. J. Agric Coop 2:40.

Journal of Dairy Science Vol. 79, No. 7, 1996