January 12, 1999

^§ The author is Professor of Resource Economics, Department of Resource Economics at the University of Massachusetts. Analyses were conducted under contract with the Northeast Dairy Compact Commission. The author is solely responsible for the content of this report and any views expressed are those of the author.

The full economic costs of production include explicit cash costs as well as implicit costs for those resources not paid a cash wage. In this study, survey data were used to determine a range of estimates for the costs of production in the New England Market area. The survey provided estimates of cash costs for those inputs that received explicit payments. Careful attention was paid to determining implicit values for those resources not paid a cash wage. Data for 143 farms in the New England states and New York were used to estimate costs of production per hundredweight of milk. The survey was conducted during the spring of 1998 and asked farmers to provide data for1996.

To arrive at final estimates of the costs of production, cash costs were estimated directly from survey data. From a sample of 143 farms, cash costs were estimated to be $14.77 on average. The median value of cash costs for the 143 was $14.58. The implicit costs of unpaid labor was estimated from hours of unpaid labor reported by the farm and a wage rate of $5.00 per hour. The average implicit value of unpaid labor was $0.37 per hundredweight, the median value was zero, which simply reflects that fact that many farms did not report using unpaid labor. The implicit costs of land owned was calculated using rental values reported by the USDA. Implicit land costs were $0.15 on average. Operator labor was valued at two different wage rates reflecting the value of operator labor on other farms ($7.18 per hour) or at other non-metro employment ($11.72 per hour). The median implicit values of operator labor ranged from $2.59 per hundredweight to $6.36 depending upon the wage used and whether all operator labor was included or only the labor reported by the primary operator. Median implicit values for net farm investment ranged from $1.99 to $2.93, depending upon the chosen rate of return.

Average estimates of the costs of production were sensitive to extreme observations in the data as shown by differences between means and medians. Therefore, only the median values for costs of production per hundredweight are presented in the table below. A range of median values is presented here to show the sensitivity of costs per hundredweight to the different assumptions used for the opportunity costs of operator labor and net farm investment.

The median costs of production in the New England Market vary a low of $18.80 to a high of $23.17. The low value of $18.80 would likely be comparable to the USDA estimates were they to include a value for operator labor. All unpaid labor is valued by the USDA at a farm wage rate (columns 2 and 4 in Table 8). Because farmers choose to stay in agriculture, even at lower wage rates, choice of the farm wage rate to value labor may be appropriate. The range of values of $18.80 - $21.06 per hundredweight would then be the preferred range for costs of production in the New England Market. One final statistic from the analysis provides an indication of the range of values for the sample of 143 farms. The inter-quartile range for the median value of $18.80 per hundredweight was $16.38 to $21.65; fifty percent of the 143 farms in the sample have costs of production with this range. Using a rate of return on investment that reflects additional risk (Aaa bonds - 7.37%), the median costs of production rises to $19.28 with an inter-quartile range of $16.80 to $22.57. Median costs of production per hundredweight were also found to vary by herd size. Using an operator wage rate of $7.18 and a rate of return for net investment of 5.47%, costs ranged from $22.34 for the smallest herd size to $17.60 for the largest herd size.

Regression analysis was used to test whether the costs of production differed across states. A set of dummy variables were created to test whether the costs of production per hundredweight for Connecticut, Massachusetts, Maine, New Hampshire, New York or Rhode Island farms were statistically different from those of Vermont. The regression results indicated suggested that the issue of extreme values was again a concern. To alleviate the effects of potential outliers in the data, the upper and lower 5% of the data were deleted. The results of the regression analysis indicated no statistical differences between the costs of production for the seven states in the sample. While the estimated parameters were positive for variables indicating cost differences for Connecticut, Massachusetts, Maine, New Hampshire and Rhode Island, these coefficients were not statistically different from zero. Thus, we conclude that the full economic costs of production are not different across the seven states included in this study.

Daniel A. Lass

University of Massachusetts

Developing accurate estimates of the costs of production requires that all resources used in production be valued. Costs of production are classified broadly into two categories: explicit (cash costs) and implicit expenses. The first category, explicit costs, includes productive resources that receive explicit payments or wages. Items in this category include purchased feeds, fertilizers, seeds, fuels, oils, utilities, and payments to hired labor as well as other familiar variable costs. In addition, costs associated with farm capital inputs or investments are included through depreciation, interest, repairs, taxes and insurance payments.

The second category of costs is more difficult to measure. Implicit costs accrue to the farm operator and the operator’s investments. For example, if the farm operator does not receive a cash wage, then the labor input of the operator should be valued at the opportunity cost of his or her time. This is a very common issue in agriculture; most farm operators act as residual claimants. The operator’s equity in capital and real estate should also be paid an opportunity cost in addition to the cash costs associated with farm investments. Rather than maintaining their investment in a dairy farm, the operator could invest in alternative enterprises or financial instruments. The operator’s equity in the farm should be paid an opportunity cost equal to the rate of return that could be earned through an alternative investment. Finally, while interest payments on the purchase of land may be captured in explicit costs, a rental value of the land owned should also be included as payment to land as a productive resource.

In the analysis below, the 1996 cost of production survey conducted in the New England Market area will be analyzed to provide estimates of the costs of milk production for the New England area. The New England Market area includes Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island, Vermont and a portion of New York. A set of descriptive statistics will first be presented to characterize the survey respondents. Then costs per hundred pounds of milk produced will be calculated from the survey data. Detailed summary measures of cash costs as reported by the survey respondents will be presented. Implicit costs will be estimated for operator and unpaid labor, farm equity and land using survey responses and secondary data. The final set of full economic costs per hundred pounds produced will be based on assumptions made for operator and unpaid labor wage rates, rates of return for farm investments and rental rates for farm land. A range of values will be presented to show the sensitivity of the estimates to the assumptions made.

The 1996 dairy cost of production survey was conducted during March of 1997. The survey was conducted by the New England Market Administrator’s office using the administrator’s list of dairy farms supplying the New England Market. A random sample of dairy farms was selected from all states supplying the New England Market, including the six New England States and New York. The surveys were returned directly to the University of New Hampshire Survey Center; a total of 273 survey forms were returned. On average, 4,019 dairy farms produced for the New England Market in 1996, with 3,891 producers remaining in December of 1996 (Market Administrator, New England Milk Market Statistics - 1996). The percentages of producers by state are as follows: Vermont 44.4%, New York 25.2%, Maine 10.1%, Massachusetts 8.0%, Connecticut 6.2%, New Hampshire 5.3%, and Rhode Island with less than 1%. The sample obtained from the survey represents about 7% of the average number of farms that supplied the New England market in 1996.

Table 1 presents a number of descriptive statistics for the survey respondents. Of the 273 respondents, 246 operated their dairy farm for the full year in 1996. Of all respondents, just over two-thirds farmed in either Vermont (43.9%) or New York (22.9%). Of the remaining New England states, Maine had the next highest share of respondents (11.1%) followed by Massachusetts (7.7%), Connecticut (7.0%), New Hampshire (4.4%) and Rhode Island (3.0%). The numbers of respondents by state match the actual distribution of producers in the New England Market area fairly closely. The distribution of producers in the survey appears to be representative of the New England Market distribution.

Table 1 also presents a number of farm characteristics. The most prevalent form of business organization was the sole proprietorship employed by 66.8% of the farms. An additional 25.9% of the farms were organized as partnerships. The most common form of record-keeping was reported to be a manual form such as farm account book or ledger (52%). An on-farm computer was used by over 16% of the farms. Mail-in record services and private bookkeepers/accountants were each used by 14-15% of the farms. The average age of the primary operator was just over 52 years and the primary operator had more than 34 years of dairy farm experience on average. Several familiar characteristics measure the scale of the average operation. Average farm size was 378 acres, about 296 acres were used for the dairy operation. The average herd size was about 98 milking cows and those cows produced an average of just over 15,000 pounds of milk annually.

A number of respondents did not farm for the full year in 1996. Our objective is to estimate the costs of production for the full year and it was not possible to determine what portion of the year those farms did produce. Thus, those farms that did not operate for the full year were dropped from the data set leaving a total of 246 observations (farms) for the analysis of production costs. A number of farms also had missing values for key data on costs and were also deleted from the sample. In the analyses of costs that follow, two subsets of the data were used: (1) a data set of 223 farms that provided information on cash costs; and (2) a data set of 143 farms that provided information on values of farm assets and liabilities as well as cash costs. Summary statistics for these two subsets are presented in Table 2. The statistics in Table 2 show that the two subsets represent the entire sample quite well. There are no significant differences in the means across these three different samples.

Costs of production are estimated in several steps as described briefly in the introduction. In the following sections, more detail is provided on the methods used to estimate different costs. The estimates obtained are presented following each description of methods. As indicated above, the goal in this study was to estimate the full economic costs of production. The explicit costs of production are estimated first, we will then turn to estimation of the implicit costs. Explicit costs of production are measured directly from survey data. Implicit costs of production use physical resource measures from the survey data, which are then valued by using secondary information to establish wages, rates of returns and rental rates. A range of values is presented for the final costs of production to illustrate the sensitivity of estimates to assumptions about the implicit values assigned to the resources.

Cash Expenses

Cash expenses are payments to factors that receive an explicit wage or payment. Cash costs are further separated into variable cash expenses and fixed cash expenses. Variable cash expenses are those costs that the farm will incur only if production takes place. For example, if the farm curtails production (shuts down in the short-run) feed, labor and fuel expenses associated with producing milk will no longer be incurred. However, fixed cash expenses will still be incurred, at least in the short-run. Fixed cash expenses are those costs that the farm must incur whether or not production takes place, such as interest payments and taxes. The cash expenses reported here are determined by summarizing the survey responses. These cash costs are then adjusted for any changes in inventories and accounts payable that were also reported by the survey respondents.

Data were collected on cash payments by the farm for more than 30 inputs used in milk production. Closely related inputs were aggregated into a smaller number of categories in order to provide a more concise summary of expenses per hundredweight of milk produced. Summary statistics for cash expenses are presented in Table 3 for the 223 farms that provided complete cash expense data. Two summary statistics are presented, the sample mean and median. The median is less sensitive to possible outliers in the data and is therefore the preferred measure of central tendency if the distribution of responses is skewed. For example, average feed costs per hundredweight of milk produced are lower than median feed costs, suggesting a distribution that is skewed to the left. A comparison of the mean and median for hired labor expenses suggests just the opposite, a distribution for hired labor that is skewed to the right.

Cash expenses in Table 3 are broadly categorized as either variable or fixed cash expenses. On average, 1996 variable cash expenses of production for these 223 farms were $9.71 per hundredweight. Adding average fixed cash expenses of $4.64 brings the average total cash expenses to $14.48 per hundredweight. Median variable cash expenses were very close to average variable cash expenses as were the mean and median for total cash expenses. Aggregation of input expenses tends to balance the distributions of individual farm expenses such as feeds and labor that may be skewed right of left.

As is typical of dairy farms, feed expenses represented the largest bill that dairy farmers faced, comprising 39% of variable cash expenses and 26% of total cash expenses, on average. Median feed expenses were 45.5% of median variable expenses and 31% of median total cash expenses. Fixed cash expenses represented 32% of total cash expenses. Capital costs were the largest portion of fixed cash expenses. Average capital expenses represented 80% of fixed cash expenses and 26% of total cash expenses. Median capital expenses were 24% of median total cash expenses suggesting the distribution was skewed by farms with large capital costs. Capital costs included depreciation, interest paid on loans, repairs and maintenance, taxes and insurance. These cash costs constitute only part of the costs of capital services. The final component of the cost of capital services, opportunity costs, will be discussed below. Fixed cash expenses represent costs that the farm operator can not avoid during short-run declines in production. However, it is not uncommon for dairy farm operators to forgo paying depreciation or replacing depreciated capital during lean years.

Accrual Adjustments

A number of items purchased can be held as inventories making it important to adjust cash expenses for inventory changes that affect the actual input amounts used for 1996 production. Survey respondents provided beginning and ending inventory values for a number of categories as detailed in the appendix. Positive increases in inventories represent inputs purchased but not used during 1996 and were deducted from cash expenses. Conversely, decreases in inventories indicate inputs remaining from 1995 purchases that were actually used in1996. Thus, decreases in inventories were included as a cost of production. An additional adjustment was also made for changes in accounts payable. Increases in accounts payable represent inputs that were obtained, but not paid for, during 1996. Increases in accounts payable are added to cash expenses; decreases in accounts payable are deducted from cash expenses. These combined adjustments also account for any inputs obtained, but neither used nor paid for during 1996. As can be seen in Table 3, accrual adjustments and accounts payable adjustments increased cash expenses by a combined $0.13 per hundredweight on average.

Table 4 summarizes cash expenses for the subset of 143 farms that also provided complete data on farm assets and liabilities. Data on farm assets and liabilities were needed to calculate opportunity costs for the operator’s investment. While the average total cash expenses were higher by about $0.30 per hundredweight for this sample of farms, median total cash expenses are very close for the two samples. As we found above for descriptive characteristics of the farms (Table 2), the sample of 143 farms appears to resemble closely the larger sample. Analyses below will utilize the sample of 143 farms that provided the data needed to calculate full economic costs of production.

Implicit costs of production were calculated for all farms providing the necessary data. The reporting of values of farm assets and liabilities proved to be a limiting factor in the number of observations available for estimation. As a result in 80 farms were dropped from the sample of 223 farms due to missing values. The remaining 143 farms were used to calculate implicit costs and derive the final estimates for the costs of production.

Implicit Labor Expenses

Two forms of labor are employed on dairy farms: (1) hired labor that is paid a cash wage, and (2) labor that does not receive a cash wage. Cash wages and the value of fringe benefits for hired labor were reported on the surveys and were included as variable cash expenses. The second category includes both farm operator labor that was not paid a cash wage and unpaid labor. Implicit costs of production are calculated for operator and unpaid labor because they represent productive resources that did not receive an explicit wage. A number of farm operators did not pay themselves a cash wage, in fact only 49 of the 143 operators received a cash wage. These two forms of farm labor were valued by applying wage rates from secondary sources to the data on hours worked obtained from the survey.

Implicit costs for farm operator labor were calculated using survey data for hours of operator labor and two alternative wage rates. New England Agricultural Statistics (1997) reported 1996 quarterly farm wages for four classes of labor: field, livestock, field & livestock, and all hired workers. A simple average of the 1996 quarterly values for all hired workers was $7.18 per hour. This wage represents an opportunity cost to the operator of working on his or her own farm rather than another farm and can be used as one value of operator labor. In using this wage rate, we assume that the opportunity cost of operator labor is best provided by alternative employment within agriculture. A second wage was based on the 1996 earnings per job in non-metro areas by state (USDA, ERS, 1998). Wage rates were imputed for each state assuming a 40 hour work week. A simple average imputed wage rate for the seven states considered in this analysis was $11.72 per hour. This alternative opportunity cost assumes the best alternative value of operator labor is outside the farm sector. Finally, a number of farms reported more than one operator. To show the sensitivity of the estimates to the hours of operator labor reported, opportunity costs of operator labor are reported for only the first operator and for all operator labor.

Many farms provided data for hours of unpaid labor. A value was calculated using a modest wage rate ($5 per hour) and the reported hours of unpaid labor. The average value of unpaid labor per farm was $3168; the average value per hundredweight was $0.37.

The implicit costs of operator labor are presented in Table 5 both on a per farm and per hundredweight produced basis. The average values per farm indicate the magnitude of annual cash wages dairy farm operators would earn were they to receive a wage for the hours they worked. The two wages used here reflect the opportunity cost of their time in agriculture and in non-metro areas in general. For example, using the number of hours reported by average farm operator (about 72 hours per week) and the non-metro wage, the primary farm operator would have earned 1996 cash wages of $43,983, a reasonable income. The differences between the means and medians suggest that once again, the results are sensitive to large positive values.

Opportunity Costs of Capital and Other Assets

Farm operators typically have significant investments in the farm that could be earning returns from alternative sources such as an alternative business investment, a portfolio of financial instruments, or a savings account. The opportunity cost of net asset values was calculated using the survey data on farm assets and liabilities and secondary data on rates of return for alternative investments. First, the value of assets for each farm was calculated for seven different categories by averaging the beginning and ending values. The net worth of the farm was calculated by deducting liabilities from the total value of farm assets. Liabilities were also calculated as the average of beginning and ending values. Several different rates of return were collected, reflecting different levels of risk. The average rate for certificates of deposit, a relatively risk free investment, and average yields on corporate bonds were collected for 1996. These rates were then applied to farm net worth to calculate three different estimates of opportunity cost. Opportunity costs of the operator’s net investment per hundredweight of milk produced are presented in Table 6. Mean and median values of opportunity costs again vary significantly suggesting the mean value is sensitive to large positive values.

Implicit Cost of Land

The final implicit expense that is included is an opportunity cost for the number of acres of farmland owned. The opportunity cost was calculated using rental values for farmland published by the USDA, National Agricultural Statistics for New England and New York. The National Agricultural Statistics Service reported that cash rent per acre for cropland in New England was $30.70 in 1996. Cash rent per acre for New York cropland was $29.00 in 1996. Pasture rents were not reported for New England in 1996; the 1995 cash rent per acre for pasture of $20.90 was used. New York rents for pasture were $14.50 in 1996, the same rate as 1995. Thus, at least in New York, rental rates for pasture were stable from 1995 to 1996 and lacking additional data the same is assumed for New England.

The implicit costs of land were calculated by first determining the number of acres owned that were used in the dairy operation. Thus, from the number of acres operated, we deducted the acreage rented, assuming that all rented acres were used in the dairy operation and that these costs were included in cash expenses. The proportion of farmland in cropland as reported by the USDA Economic Research Service was then used to separate the land owned into cropland and pasture. The rental rates for cropland were applied to the proportion estimated to be cropland; the remaining land was valued at the rental rate for pasture. The average implicit cost of land owned was calculated to be $0.21 per hundredweight of milk produced. The median value was lower at $0.12 per hundredweight of milk produced. These values are the estimates of implicit costs of land that will be included in the final cost of production estimates.

The full economic costs of production include explicit cash costs and implicit costs for those resources not paid a cash wage. To arrive at final estimates of the costs of production, results for those factors discussed above were summed. The implicit costs of operator labor (Table 5), unpaid labor, net farm investment (Table 6) and land were added to the total cash costs reported in Table 4. Some operators did pay themselves cash wages; these wages were deducted before adding the opportunity costs of operator labor.

The final estimates of average costs of production per hundredweight are presented in Table 7. A range of values is shown to illustrate the effects of different assumptions about the opportunity costs for operator labor and capital on costs of production. Four different valuations of operator labor were considered: (1) only the primary farm operator’s time valued at $7.18; (2) only the primary farm operator’s time valued at $11.72; (3) all reported operator labor valued at $7.18; and (4) all reported operator labor valued at $11.72. There were three different rates of return used to calculate the opportunity cost of net farm investment: (1) a CD rate of 5.47%; (2) the Aaa bond rate of 7.37%; and (3) the Baa bond rate of 8.05%. The low average value of $20.38 per hundredweight in Table 7 was represents a combination of cash expenses less payments to the operator and family members ($14.77 less $0.60 on average) plus unpaid labor ($0.37), implicit land costs ($0.21), operator labor at $7.18 per hour ($3.17) and opportunity costs of capital at 5.47% ($2.50). These calculations were completed for each farm and the averages computed for the 143 farms in the sample.

The average final cost estimates presented in Table 7 are sensitive to extreme values as discussed throughout the text. Median costs per hundredweight of milk produced are presented in Table 8; median values are not as sensitive to extreme values. A range of median values is presented to again show the sensitivity of costs per hundredweight to the different assumptions used for the opportunity costs of operator labor and net farm investment.

The median final costs of production in the New England Market vary a low of $18.80 to a high of $23.17. The low value of $18.80 would likely be comparable to the USDA estimates were they to include a value for operator labor. All unpaid labor is valued by the USDA at a farm wage rate (columns 2 and 4 in Table 8). Because farmers choose to stay in agriculture, even at lower wage rates, choice of the farm wage rate to value labor may be appropriate. The range of values of $18.80 - $21.06 per hundredweight would then be the preferred range for costs of production in the New England Market. One final statistic from the analysis provides an indication of the range of values for the sample of 143 farms. The inter-quartile range for the median value of $18.80 per hundredweight was $16.38 to $21.65; fifty percent of the 143 farms in the sample have costs of production with this range. Using a rate of return on investment that reflects additional risk (Aaa bonds - 7.37%), the median costs of production rises to $19.28 with an inter-quartile range of $16.80 to $22.57.

Costs of production also vary by herd size. Categories were created using quartiles for number of cows milking. Median costs of production per hundredweight are presented for 4 different herd sizes: 47 cows or fewer; between 48 and 65 cows; between 66 and 90 cows; and greater than 90 cows. Costs of production per hundredweight decreases as herd size increases. Using a wage rate of $7.18 and rate of return on investment of 5.47%, farms in he largest herd size category had median costs per hundredweight of $17.60 compared to a median of $22.34 for the smallest herd size category. Similar results hold for the different wage rates and rates of return on investment. This result was also confirmed in subsequent regression analyses where milk produced was used as an explanatory variable for the costs of production.

Differences in the costs of production across states can be analyzed in several ways. One approach is to compute means for each state, and then test whether the state meannsusing the appropriate univariate statistical tests. There are two limitations to this approach. First, given the analysis above, the average values appear sensitive to extreme values, or what may be statistical outliers. Second, there are a number of factors that may lead to differences in costs per hundredweight, including size of the dairy operation, measures of efficiency, and human capital measures for the farm operator. In order to capture these potential effects, a multivariate approach to estimating differences across states should be followed. In the analysis below, multiple regression analysis is used to estimated differences across states in the costs per hundredweight of milk produced.

The first task in testing for differences in the cost of production across states was to create a set of dummy variables, or binary variables, indicating the state where the farm is located. The dummy variables were then included in multiple regression models and estimated coefficients of the dummy variables represent difference between costs of production in those states and the state chosen as the "base." Vermont, the state with the greatest number of farms was chosen as the "base" for each of the models estimated. Thus, the multiple regression models contained dummy variables for: CT, MA, ME, NH, NY and RI.

It is important to include other factors that may affect the costs of production. Leaving important factors out of the model result in biased estimates. Previous analyses have found that farm size or total output, efficiency, and operator human capital affect the costs of production. Farm size was measured by the total amount of milk produced. The shape of the relationship between costs and output will vary depending upon the measure of costs that is considered. Average variable costs of production and average total costs of production should initially decline and then begin to rise as output increases. To capture this typical "U-shaped" relationship, total milk produced was included in a quadratic form. Measuring efficiency is difficult and data intensive. However, one measure that indicates relative efficiency across farms and the genetic potential of the herd is output per cow.

Operator human capital is measured by the level of education and years of experience for the primary operator. Education level was measured by dummy variables indicating whether the operator had a high school degree (HS), a two-year college degree (AS), or a 4-year college degree (BS) or greater. The number of years experience as the farm operator and the total number of years experience in dairy farming were also included as measures of human capital. We anticipate that these human capital variables will negatively affect the costs per hundred weight.

Regression models were estimated for average variable costs, average fixed cash costs, average total cash cost, and average total costs of production (full economic costs) for the sample of 143 dairy farms. Statistically significant differences were found in the average variable cash costs of production for Connecticut, Massachusetts and Rhode Island. Rhode Island was found to have higher average fixed costs of production. Average total economic costs of production for Connecticut, Massachusetts, Rhode Island and Maine were found to be statistically greater than the costs of production for Vermont.

When the average costs of production were presented above, concerns were raised about the effects of extreme values on the statistical results. Use of the median to summarize the costs of production was favored. The same problem was found upon further analysis of the regression results; a number of farms appeared to be statistical outliers. To correct for the affect of such extreme values, the upper and lower five percent of the data were deleted from the data set to delete potential outliers. These new regression results should be more consistent with use of the median to describe the economic costs of production.

The data used in the final regression results are summarized in the Appendix, Table A.3. The median costs of production as shown in Table 8 are unaffected by such a "trimming" of the data. Again, Connecticut and Massachusetts dairy farms were found to have statistically greater average variable cash costs of production as compared to Vermont farms. New York dairy farms had statistically greater average fixed cash costs of production. However, none of the states had statistically different full economic costs of production, the measure of primary interest for this study. The factors that had the greatest affect on the full economic costs of production were farm size or output level, education and total dairy farm experience. Those operators who had a BS degree had lower costs of production by $2.96 per hundredweight, on average. Each additional year of dairy farm experience lowered the costs of production by an additional 5.3¢ per hundredweight, on average.