MSU RETURNING MATERIALS: Place in book drop to nan/mugs remove this checkout from .-,—. your record. FINES will be charged if book is returned after the date stamped below. - ‘7) , , f‘"-.- , : -.-.'¢ km:- .‘...4_»- .‘ ,'h z a ,i If, ~‘f 3,, ‘1... ‘-'=' -1 PROTOCOL FOR THE COST—BENEFIT ANALYSIS OF DAIRY CATTLE HEALTH MANAGEMENT BY Paul A. Cummins A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Agricultural Economics 1983 ABSTRACT PROTOCOL FOR THE COST-BENEFIT ANALYSIS OF DAIRY CATTLE HEALTH MANAGEMENT by Paul A. Cummins This study establishes a protocol for the cost-benefit analysis of dairy cattle health management which will be utilized within the Food Animal Health Resource Management System (FAHRMX) at Michigan State University. The goal is for the data storage and processing capabilities of microcomputers to be exploited for the rigorous economic analysis of specific disease con- trol procedures on individual commercial dairy farms. The data requirements and modeling difficulties that must be overcome for such analysis are dis- cussed. A single equation multivariate linear model of milk production, based ‘ on data available previous to FAHRMX, is used to demonstrate how inclusion of culled cows helps correct for the high positive parameter estimate expected for cystic ovaries. Future improvements depend on the identification of.a set of simultaneous equations. Results of a questionnaire concerning farm infrastructure as it relates to dairy cattle health care are also presented. The cow is of the bovine ilk; one end is moo, the other, milk. The Cow, by Ogden Nash ii ACKNOWLEDGEMENTS In the true spirit of a university this Master's project has received support from several people in varied disciplines: Dr. Paul Bartlett and Dr. Edward Mather in the Department of Large Animal Surgery and Medicine, Dr. Clyde Anderson in the Animal Science Department, Dr. Sherrill Nott and Dr. J. Roy Black in the Department of Agricultural Economics. All of these people have made intellectual contributions to my work. Special thanks go to Dr. Mather, Dr. Anderson, and Dr. Nott for consistently providing moral support as well. I am indebted to the W.K. Kellogg Foundation which funded this pro- ject, thus assuring that my debt to others mentioned here could be strictly non-monetary. Mr. A1 Thelen and his staff at the Dairy Herd Improvement Association welcomed my research and facilitated utilization of their records. I would also like to thank all the secretaries who so often bore the brunt of my advisors' eccentricities and my own impatience: Ms. Linda Peters, Ms. Debbie Greer, Ms. Diane Leslie, Ms. Linda Smith, Ms. Denise Arnold, and Ms. Debra Andrews. My brother, Dr. Mark Cummins, helped edit the manuscript. Appreciation to other family and friends is best expressed in my daily life. To all the dairy farmers and veterinarians that made my work realistic, exciting and possible I offer this thesis in hopes that they will one day be served by it. iii TABLE OF CONTENTS List of Tables List of Figures A. D. Page vii ix Chapter One. Introduction 1 The goal of efficient resource use 1 1. Both monetary and non-monetary objectives are relevant l 2. More monetary accounting information helps with both monetary and non-monetary decision making 1 Great potential gains to be had by improving animal health management are largely hidden from dairy farmers 2 1. Detection of hidden losses requires careful monitoring of performance 3 2. FAHRMX is now developing a means to carefully monitor dairy herds' health management performance 3 a. FAHRMX data collection 3 b. FAHRMX data formatting 4 c. FAHRMX data processing 5 Comparative medicine and the farmer's resource-use decisions are equally well served by cost-benefit analysis 5 Scope of thesis 6 _ Hkn/PH Chapter Two. The Economic Evaluation of Dairy Cattle~Herd Management: A Review of the Literature 8 Introduction 8 Estimating Losses Due to Mastitis 9 Estimating Losses Due to Reproductive Health Management Problems 10 Measuring the Value of Dairy Cattle Health Care 12 l. The Rise of Intensive Preventfiive Care 12 13 a. Hershfler et a1. (1964) iv E. Grunsell et a1. (1969) Barfoot et a1. (1971) Poterfield and Heider (1980) McCauley (1974) send Computer Applications to Dairy Cattle Health Management Chapter Three. Data and Methods A. 8. Introduction Data Sources 1. 2. 3. Description of Pilot Herds Questionnaire a. Purpose and Description b. Sample Size Limitations Retrospective Data File Data Applications 1. General Concepts of Cost-Benefit Analysis a. Introduction b. The Difference Between Financial and Economic Analysis A Basic Tool of Cost-Benefit Analysis: The Partial Budget Itemizing the Costs and Benefits of Disease Control for Dairy Cattle Examples of Partial Budgeting to Evaluate Disease Control Procedures a. Introduction b. Examples A Model to Estimate Reduced Lactating Potential Due to Disease a. Measuring Production b. Scope of the Ideal Model c. Format of the Statistical Mbdel Using Retrospective Data d. Possibilities for Future Improvement of the Model: Simultaneous Equations e. A Point of Clarification: Measuring Reduced Lacta- ting Potential in Animals Treated for Disease f. Conclusion to Methods 16 17 19 19 23 26 26 28 28 31 31 31 33 36 36 36 37 37 39 51 51 S4 55 S6 62 62 68 69 70 Chapter Four. Results 71 A. Introduction 71 B. Questionnaire Results 71 C. Comparison of Sample Disease Data to That of Other Studies 76 D. Characteristics of Culled Cows 76 E. Correlation Analyses Compared: Sample With and Without Culled Cows 80 F. Regression Results Including Indices of Genetic Potential 87 G. Summary and Conclusions 87 List of References 90 Appendix I 93 Appendix 2 94 Appendix 3 104 Appendix 4 115 Appendix 5 132 Appendix 6 133 Appendix 7 134 vi Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table 10. 11. 12. 13. 14. 15. LIST OF TABLES Value of a Reduction in Calving Interval (CI) Over Different Ranges Assuming Reduction of CI Sole Cause of Production Increase Some Advantages and Disadvantages of the Herd Health Program Evaluation Studies Selected Characteristics of Holstein Herds in FAHRMX for the Year Prior to Study Initiation (1980) Itemization of the Costs and Benefits of Disease Control for Dairy Cattle Example of the Importance of Using Present Values Herd Size, 12 of 23 FAHRMX Participants, 1980 Milk Production Class, 12 of 23 FAHRMX Participants, 1980 Housing for Lactating Cows, 12 of 23 FAHRMX Parti- cipants, 1982 Location of Milking, 12 of 23 FAHRMX Participants, 1982 Housing for Dry Cows, 12 of 23 FAHRMX Participants, 1982 Location of Calving, 12 of 23 FAHRMX Participants, 1982 Extent of Disease Recording Before the Utilization of FAHRMX, 12 of 23 FAHRMX Participants, 1979-1981 Milk Fever, Percentage of Farmer Treatment, 12 of 23 FAHRMX Participants, 1983 Retained Placenta, Percentage of Farmer Treatment, 12 of 23 FAHRMX Participants, 1982 Percent Culling Per Year, Eight Pilot Herds in FAHRMX Retrospective Sample vii 16 21 29 39 53 71 72 72 72 73 73 75 75 75 81 Table Table Table Table Table 16. 17. 18. 19. 20. Reasons for Dairy Cattle Replacement 82 Culling Summary, Eight Pilot Herds in FAHRMX Retrospec- tive Sample 83 Single Equation Run Without Culled Cows 84 Single Equation Run With Culled Cows Included 86 Single Equation Run Including Indices of Genetic Potential 88 viii Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure 10. ll. 12. 13. 14. LIST OF FIGURES Location of FAHRMX Participants and Sources of Sample Data, 1982 Culling Behavior in the Absence of Disease Problems Severe Enough to Warrant Culling Culling in the Presence of Disease Problems Severe Enough to Warrant Culling: Assuming Every Tenth Cow Culled for Disease Reasons Loss of Production Potential Caused by Loss of Culling Options Definition of Optimal Calving Interval (CI) as per Louca and Legates (1968) Possible Effect of Disease on Lactating Potential "Dip" in Milk Production Undetected by Monthly Testing Procedures "Dip" in Milk Production Overestimated by Monthly Testing Procedure Use of 305 Day Projection Factors to Detect Reduced Lactating Potential Illustration of Culling Bias Days in Milk at Time of First Treatment for Metritis, Eight Pilot Herds Days in Milk at Time of First Treatment for Cystic Ovaries, Eight Pilot Herds Comparison of Hansen's and Shanks Reproductive Dis- order Cost Distribution to Cummins' Combined Incidence «for Data for Cystic Ovaries and Metritis from Eight Pilot Herds Distribution of Culls by Production Level, Eight Pilot Herds ix 43 44 46 49 57 59 6O 61 66 77 78 79 135 Chapter One Introduction A. The goal of efficient resource use. Economically efficient use of resources occurs when each resource makes its maximum contribution to predetermined objectives. The goal of economically efficient employment of resources is obtainable only within the limits of known applications. Resource owners, dairy farmers for our purposes, make estimates of the expected return from investment of their resources in each known application (expected return = probability of payoff multiplied by the payoff). They weigh the expected return of each investment against the other known alternatives when deciding how to best utilize their time, money, and land. The dairy farmer's estimate of expected return may be more or less explicit depending on the amount of accounting information available. 1. Both monetary and non-monetary objectives are relevant. The maximization of monetary profit is one common objective. However, there are non-monetary objectives relevant to dairy farming. Farmers may have favorite animals that receive care unjustifiable in terms of monetary gain alone. In some cases, their purchase of ever more expensive equipment, including new computer technology, may be primarily motivated by social status considerations (see Appendix 1). Investments made for non-monetary reasons have consequences that can be measured in dollar terms, however. in fact, all the decisions concerning resource use on the farm have consequences that can be given a dollar value. 2. More monetary accounting information helps with both monetary and non-monetary decision making. When a dairy farmer invests in a given venture, the gain to be had from the next best alternative is sacrificed. \Vhen farmers' primary objective is monetary profit, they would like to know in what use their resources are likely to yield the highest monetary return. Presumably, farmers would also like to know the amount of money they forego when they utilize resources to meet non-monetary objec- tives. The fulfillment of both monetary and non-monetary objectives would be facilitated by more clearly delineating the monetary consequences of farm resource use. With monetary accounting information, objective weights can be given to subjective decisions. This means that fewer investment "mistakes" will be made when more monetary accounting information is available. Therefore, regardless of the farmers' objectives, they will benefit from knowing more precisely what the monetary value of their resources are in different applications. B. Great potential gains to be had by improving animal health management are largely hidden from dairy farmers. When a calf or cow dies, the farmer recognizes the loss of all the milk and offspring that would have been produced by the animal had it not died. Other losses of productive potential are not so obvious. For example, when productivity is impaired by subclinical mastitis or infertility, nothing that physically existed is taken away from the farmer. What is lost in these instances is Etential. Because losses from subclinical mastitis and infertility are difficult to recognize and quantify, investment in mastitis and infertility control are among the farmer's least well-known alternatives. For these same reasons, many veterinary treat- ments for mastitis and infertility remain controversial. The potential for profit through increased attention to dairy animal health management is evidenced by the enormous dollar losses attributed to mastitis and infertility. Blosser (1979) estimated that $1.3 billion was lost due to mastitis in U.S. dairy cattle in 1976. He attributed 69 percent of this loss to subclinical mastitis. Meyer (1953) reported that impaired fertility causes a total annual loss of $800 million in 0.5. cattle. Adjusting only for inflation, this latter estimate would have to be more than tripled to bring it up to date.y Although the validity 1/ - All price adjustments for inflation made using the change in consumer price index for all items (Statistical Abstract of the United States, 1981). of these estimates may be questioned, nonetheless they indicate the magnitude of these two health problems. 1. Detection of hidden losses requires careful monitoring of performance. Detection of lost production potential, such as that caused by mastitis and infertility in dairy cattle, requires careful monitoring of a herd's performance. Health and production records must be combined and compared over time. The data requirements for such monitoring are significant. Computer technology has made the cost of extensive data storage and processing very low. A research project at Michigan State University called the Food Animal Health Resource Management System (FAHRMX) has demonstrated that detailed animal health management data can be collected and computerized with minimal farmer effort (see Appendix 2 for FAHRMX grant proposal). 2. FAHRMX is now developing a means to carefully monitor dairy herds' health management performance. FAHRMX is an experiment initially involving 24 dairy farms served by 5 veterinary practices throughout Michigan. There are three stages of the system's function: data collection, data formatting, and data processing. a. FAHRMX data collection. Data currently being amassed by FAHRMX include records of all veterinarian and farmer delivered health care: vaccinations, disease treatments, calving dates, results of reproductive exams, etc. When participating farmers or veterinarians treat an animal, they record, in their own words: the date, the animal ID, and the action taken. The action taken may include the time spent in treatment, the dosage of drug administered, and any other details they deem pertinent. Their script is translated into codes and entered onto microcomputers by technicians at the five veterinary offices. Other data are collected for FAHRMX through a questionnaire administered to each dairy farm concerning housing, milking, and health treatment facilities, among other things (see Appendix 3). These data will be most useful for interfarm comparisons of performance. Communication between the five microcomputers and the mainframe com- puter at Michigan State University will allow for the aggregation of data from the five veterinary practices and make interfarm comparisons more meaningful. Such communication will also permit the exchange of information between the Dairy Herd Improvement Association (DHIA), which is a milk production monitoring service, and FAHRMX. Communication with DHIA will not only allow immediate access to milk production information but will also provide genetic data from DHIA's semiannual inventory reports. In conclusion, the total data collecting ability of F AHRMX encompasses farm infrastructure, animal health histories, labor and drug expense for health care, genetic information, and milk production. b. FAHRMX data formatting. There are a number of useful features of FAHRMX that simply involve organizing, or formatting, the data mentioned above. First, FAHRMX design allows for the easy retrieval of health, and eventually, milk production data on individual animals. This feature is important when, for example, an animal's health history must be considered before administering additional treatment. Second, FAHRMX software calculates whole-herd and whole-system statistics such as disease incidence figures. This provides some basis of comparison between herds, which will help farmers decide what their major animal health management problems are. Third, the existence of the data and the computer's quick searching ability are exploited by FAHRMX to raise "flags" or notices of upcoming necessary action like reproductive exams or vaccinations. The above three capabilities are also combined when, for example, the results from previous reproductive exams are displayed for all animals that are currently due for examinations. c. FAHRMX data processing. Although there are many benefits of merely reorganizing and aggregating FAHRMX's detailed animal health-related data, the most exciting opportunities presented by the data are their use for comparative medical purposes. It has never before been possible to measure, on a continuing basis, the effects of different disease control procedures on the milk producing potential of commercial dairy herds. C. Comparative medicine and the farmer's resource-use decisions are equally well served by cost-benefit analysis. Veterinarians and farmers alike are interested in which treatments and procedures are most effective in reducing cost and improving the productivity of their herds. The delineation of the costs and benefits of each health control technique would reveal the value of expenditure in each procedure to both farmers and veterinarians. Sudi cost-benefit analysis would be new to veterinary medicine. Farmers could see the effectiveness of different health management procedures working within their own resource constraints. Many of the losses from health management problems would no longer be hidden. This would work to the veterinarians' advantage because they would no longer just have research herd results with which to convince a farmer of the benefits of a new technique. A discriminating cost- benefit analysis system for dairy herd health would also provide a check on the recommendations that veterinarians make. The assumption was made earlier that further investment in dairy cattle health care merits consideration. A cost-benefit analysis system for animal health management would show where profitable investments exist within animal health management. Profitable investments in this area certainly exist. However, this does not mean that it is in the farmer's best interest to pour all available funds into animal health management. Even with a cost-benefit analysis system for dairy cattle health management, farmers would still have to decide if there are more profitable investments outside the scope of health management. For now, the development of a cost-benefit analysis system for dairy cattle health management is a big enough task. It is best to start where the greatest benefits are expected, but the need for economic decision aids for other aspects of dairy farming should be recognized. Once a cost-benefit analysis system is operational for dairy cattle health management, it could be expanded to include other farm enterprises. D. Scope of Thesis The purpose of this thesis is to develop a protocol for the cost-benefit analysis of dairy cattle health management. The economic analysis of dairy cattle health management has received much attention in the literature and has been frequently misunderstood, as evidenced in the literature review (Chapter 2). In all fairness, the rigor of previous studies has suffered from the lack of detailed health- related data, and the present study was undertaken in response to the new data capabilities outlined earlier. But the persistence of much used and very question- able estimates of losses due to disease problems is disturbing. If unsubstantiated estimates of economic losses caused by disease in dairy cattle are continually quoted because of a lack of understanding of economic analysis, then a clear presentation of the requirements for the economic analysis of dairy cattle health management is long overdue. Cost-benefit analysis is simply a form of economic analysis, or an economic model. A substantial part of this thesis is concerned with the method of cost-benefit analysis as it pertains to dairy cattle health manage- ment. The accounting tool used for cost-benefit analysis is the partial budget. The standard partial budgeting procedure includes a list of factors which reduces profit (costs) by either increasing costs or reducing income, and a list of factors which increases profit (benefits) by either increasing income or reducing costs (Harsh et al., 1981). Creating an economic model for dairy cattle disease control consists primarily in identifying relevant costs and benefits, and organizing them as is done in Chapter 3. Once it is clear how the costs and benefits will be used to aid in decision making, the next problem is estimating them. Estimating reduced lactating potential due to disease probably is most difficult because lactating potential is affected by many factors, and is itself difficult to estimate. Chapter 3 also contains a statistical model which is expected to estimate reduced lactating potential due to cystic ovaries and metritis (because these were the only diseases for which data were available retrospectively) better than has been done previous- ly. However, questions about the specification of .the model are raised, and suggestions are made as to how the model can be improved using current FAHRMX data. Chapter Two The Economic Evaluation of Dairy Cattle Health Management: A Review of the Literature A. Introduction A critical review of the available literature on dairy cattle health manage- ment reveals a growing interest, over the last two decades, in "economic" analysis. This is a function of the increasing importance given to less conspicuous effects of disease and lack of attention to animal health management. Such "hidden"reffects include production losses due to subclinical mastitis and suboptimal breeding performance. The purpose of these "economic" studies has either been to stress the need for additional research expenditure or to "prove" the value of regular veterinary visits emphasizing reproductive herd health. Estimates of industry-wide losses due to mastitis and infertility are based on little more than guesses and should therefore be viewed with scepticism. Measures of the value of regular veterinary visits or programmed herd health have suffered from the inability to isolate the most profitable features of these programs. Also, by comparing the farms' performance before and after the new program's inception, some investigators failed to correct for performance trends which were independent of the new program. In most cases, estimates of the effects of health‘management programs do not account for individual dairy farms' resource constraints. Despite these difficulties, the magnitude of the returns possible from increased attention to fertility and udder health have been demon- strated. Progress in measuring the value of dairy cattle health care depends largely on understanding the relationship between health care, disease, and milk production. Recent evidence shows a positive correlation between milk production and some diseases. Whether this correlation is spurious or not remains to be proven. All previous studies on the economics of dairy cattle health management have been limited by lad< of detailed data. Computer-facilitated data collection will allow for more rigorous analysis. B. Estimating Losses Due to Mastitis Numbers are often assumed to be magically endowed with objectivity. Once a numerical estimate is made, the subjective steps in making the estimate may be forgotten. In the case of estimating the cost of mastitis and infertility in dairy cattle, the tenuous nature of estimates made to date should be recognized. In a 1979 article, Blosser reviewed the literature concerning economic loss due to mastitis in dairy cattle in the U.S. and other countries. His goal was to emphasize the need for further research by demonstrating the magnitude of loss due to mastitis. His estimate of $1.3 billion lost in the U.S. in 1976 ($2.6 billion in 1981 dollars) is based on an aggregate of estimates made by one person from each of 33 states whom he lets "represent" 86 percent of U.S. dairy cattle. The "representatives" were asked what the magnitude of losses was in their respective states, which collectively contained 86 percent of U.S. dairy cattle. Many of these statewide estimates were based indirectly on research which related reduced milk production to results of California Mastitis Test (CMT) scores (Janzen, 1970; Forster et al., 1967; Natzke et al., 1965; Philpot, 1967). This research attributed a percentage loss of milk production to a standardized mastitis test score. With some idea of the incidence and severity of mastitis (i.e., percent of dairy cattle- showing CMT T, l, 2, or 3), the investigator could estimate reduced milk production. Unfortunately, such incidence data is not yet available for large populations. Therefore, the estimates aggregated by Blosser are little more than impressions. Research is now being carried out to relate CMT results to the Dairy Herd Improvement Association's (DHIA) method of counting somatic cells (Kirk, 1982). DHIA's Somatic Cell Count (SCC) is done electronically and requires minimal 10 additional effort during monthly milk testing. If DHIA's SCC can be reliably associated with CMT scores, then DHIA's SCC can be used to estimate reduced milk production due to mastitis instead of CMT scores. Success in this area of research would mean that data concerning the incidence and severity of mastitis could be as widespread as DHIA's network. C. Estimating Losses Due to Reproductive Health Management Problems Tracing the origins of estimates of losses due to reproductive health management problems also proves difficult. In a 1964 article, Hershler and co- authors made reference to an estimated loss of $800 million (2.3 billion 1981 dollars) due to impaired fertility in all U.S. cattle. The reference is to a paper by Meyer (1953) in which he simply listed a figure given to him by a friend. It is difficult to have much confidence in this estimate when none of the details of its calculation are known. In order to estimate the monetary benefit of reducing calving intervals through a herd health management program emphasizing reproductive efficiency, Hershler et a1. (1964) used a figure from Haller (1957) of $1.66 lost for each day beyond a lZ-month calving interval (CI). Assuming that a lZ-month CI is optimal, Haller determined from a New York survey that, "each month's delay in rebreeding means a $45 to $50 loss in production and maintenance." He attributed $20/ month for maintenance and the balance to lost production. Converting the $50 figure to days, this came out to $.66/day for maintenance and $1.00/day due to lost production. It is necessary to point out that the only real loss was due to unrealized milk production. The maintenance cost must have been paid regardless of whether the cow was pregnant or not. Assuming for simplicity that the maintenance cost was $.66 regardless of reproductive status, and that $1.00 worth of milk could be gained for each day that a CI was reduced (to a limit of 365 days), then for each day that a calving interval was extended beyond 365 days, only $1.00 11 was lost, not $1.66 as one might assume from reading Haller. In other words, $1.00 could not have been gained without losing $.66. If there is no way to have gained $1.66/day even with 365-day calving intervals, how could $1.66/day be lost when calving intervals were longer than 365 days? Louca and Legates (1968) were aware of the tenuous nature of previous estimates of losses due to extended C15, and cleared up much of the ambiguity. Louca and Legates rigorously studied the effect of "days open," defined as "the interval between parturition and successful mating," on milk production. They agreed that the length of CI provides much of the same information as days open. They found that days open are not uniformly expensive for all lactations. Each additional day open in first lactation Holstein cows was associated with an average of 1.16 kg. (2.6 lbs.) less milk per lactation period. For cows in their second and third lactations, the corresponding figures were 3.58 kg. (8.0 lbs.) and 3.68 kg. (8.2 lbs.), respectively. A reduction of 8.0 lbs. of milk represented a loss of $1.07 in 1982 (8 lbs. * $13.42/cwt. = $1.07). This estimate did not include adjustment for the reduced calving rate, with which Louca and Legates were also concerned. Estimating the cost due to a reduction in the number of calves born per year because of extended CI depends largely on the value of the calves that were not born. This rather complex problem is discussed in Chapter Three. Louca and Legates‘ results also support previous evidence that gestation does not significantly affect milk production until after the first 210 days of lactation. This suggests that the lowered milk production brought about by delayed breeding only appears after day 210. Louca and Legates concluded that lifetime milk production could be maxi- mized by keeping days open to a minimum. They suggested a 13-month CI for first calvers and a CI as short as possible for older cows. They acknowledged that the 12 limit on the minimum length of CI is the sum of a 280-day gestation period and the 26-80 days-Z, after calving required for insemination to be most successful. Their research could be improved by accounting for the reasons for extended CI. Research by Erb et a1. (1981) and results reported in this thesis show a positive association between cystic ovaries, which lengthens CI, and high production in cows. This suggests that cows with cystic ovaries make longer CIs look better because they raise the average production of cows with longer CIs. Without proper adjustment for the effects of the disease, the costs of lengthened CI, in terms of reduced production, may be misinterpreted. D. Measuring the Value of Dairy Cattle Health Care 1. The Rise of Intensive Preventive Care In the past 20 or 30 years, there has been a shift in veterinary medicine from strictly emergency service to intensive preventive care for dairy herds. By controlling the most detrimental contagious diseases and by overcoming area mineral deficiencies, veterinary science has been a crucial factor in the trend to greater herd size (Morris and Blood, 1969). With more intensive animal production, veterinary medicine has become more intensive.’ Greater emphasis has been put on management problems such as improving reproductive efficiency. A number of studies have encouraged the practice of intensive preventive veterinary medicine for dairy cattle by estimating high returns from its application. Although these studies show that preventive programs can be profitable, especially through improving reproductive performance, they could be further improved by more attention to detail. A common feature of past studies is that they calculated the value of a whole program. This made it impossible to isolate the most profitable components of a Zl26—80 days is the range of estimates they cited from other researchers. 13 program. In some studies, the final calculations were not adjusted for performance trends that existed prior to a program's inception. This usually exaggerated the value of a new program. Finally, most of the evaluation procedures ignored the increased feed, labor, and equipment costs required to increase milk production. Associated with this problem was the need to take each farm's resource constraints into account. Because resources differ between farms, there is not one optimal health management procedure for all farms (Morris and Blood, 1969). a. Hershler et a1. (1964) To measure the "economic impact" of a fertility control and herd manage- ment program on one U.S. dairy farm, Hershler et al. (1964) estimated the econom- ic benefit of reducing average days open and reducing the age at first calving for heifers. The herd in this study was maintained at 55 Guernsey cows. It was visited monthly for reproductive examinations over a three-year period. Average calving intervals were reduced by four days in the first year (433 to 429), 40 days in the second year (429 to 389), and 3 more days by the end of the third year (389 to 386)., Hershler multiplied the cumulative average reduction for each year by the number of animals (55) and added these numbers to get 220 + 2,420 + 2,585 = 5,225. This represents the total number of open days saved over the three-year period. Next, he multiplied by Haller's (1953) estimate of $1.66 saved per day that a calving interval is reduced (to a limit of 365 days) which yields $8,674. Hershler called $8,670 the "anticipated increase in income" from the control program. (For some unknown reason, $8,599 appeared in the summary instead of $8,670.) For reasons already stated, the relevant part of Haller's estimate was the production loss. Therefore, if Hershler used Haller's results for anticipating increase in gross income, he would obtain $5,225, not $8,670. To calculate the actual three-year gain in income, Hershler began with the increase of 371,195 pounds of milk which was realized over the three years. At the 14 1964 price of $5.17/cwt., this milk was worth $19,200. Hershler added $2,250 to this figure which he said represented the maintenance costs saved by breeding heifers younger. This $2,250 addition was invalid. The increase in milk production accounted for all the improvements made in the herd, including the benefit of getting heifers bred younger. The only exception is that with shorter calving intervals more calves were produced per year. The increase in milk production did not account for the increased sales of replacement stock (herd maintained at 55 cows). When the average calving interval was 433 days, 46 calves were produced per year (365/433 * 55 = 46). With an average calving interval of 386 days, 52 calves per year were produced (365/386 * 55 = 52). The inclusion of a non-existent reduction in maintenance costs exaggerated Hershler's estimate of returns. He acknowledged that his estimate did not include increased sales of replacements. Inclusion of this benefit (six more calves per year) suggests that returns exceeded $19,200. But Hershler did not account for some important costs which were associated with increased milk production, such as increased feed, labor, and equipment expenses which vary among farms. Delineation of these costs required more detailed data than wgvailable. Because Hershler used Haller's estimate, it is interesting to compare their results. Assuming that the increase in milk production could be attributed solely to reduction of days open, and that other costs and benefits balance out, then a total reduction of days open by 5,225 days was worth $19,200. This made each day's reduction worth $3.67 in 1964 or $9.53 in 1982 (371,195 * $13.42/cwt.)/ 5,225 .-. $9.53 . Single estimates of the potential production lost due to extended calving intervals are only meaningful over a given range. Each farm has a different average calving interval and can therefore expect different results from a fertility control program. 15 Referring to more of Hershler's data, it can be shown that each day's change in calving interval (days open) did not have a constant value. Comparing the information he gave about change in calving interval with the change in average annual milk production per cow, we have the following: (1) At the end of the first year of the program, milk production was 8,000 lbs. milk/cow/year and the average C1 was 429 days. (2) The second year figures were 8,500 lbs. and 389 days. (3) The last year's figures were 10,000 lbs. and 386 days. This means that between the first and second years average milk production went up by 500 lbs./yr. (8,500 ~ 8,000) while the Cl decreased by 40 days (429 - 389 = 40). If we assume that the reduced C1 is solely responsible for the increase in milk production, then we find that each day's reduction in CI was worth 12.5 lbs. milk over this range (500 lbs./ 40 days = 12.5 lbs./day). In 1964, 12.5 lbs. milk was worth $.65 (12.5 lbs. * $5.17/cwt. = $.65), or $1.68 in 1982 (12.5 lbs. * $13.42/cwt. = $1.68). Comparing the change between the second and third years, we find that a reduction in CI of three days (389 - 386 = 3) yielded a 1,500 lb. increase in annual milk production per cow (10,000 - 8,500 = 1,500). Again, assuming the reduced CI was totally responsible for the increased milk production, then each day's reduction in CI was worth 500 lbs. of milk over this range. In 1964, 500 lbs. of milk was worth $25.85 (500 lbs. * $5.17/cwt. = $25.85), or $67.10 in 1982 (500 lbs. * $13.42/cwt. = $67.10). These results are summarized in Table 1. 16 Table 1. Value of a Reduction in Calving Interval (CI) Over Different Ranges Assuming Reduction of Cl Sole Cause of Production Increase Average Average Change Change in Value/ Day, Value/Day, in C1 Production 1964 1982 Source (Days) (Pounds) ($5.17/cwt.) ($13.42/cwt.) Haller (1957) ? ? $ 1.00 Hershler et a1. (1964), Year 2 no 500 S .65 S 1.68 .-/ Hershler et a1. (1964), Year 3 3 500 $25.85 $67.10 With such a drastic difference in benefit from a day's reduction in CI, it is clearly misleading to average the benefit over a long period. Hershler's study utilized only one farm. The lack of a control group means that increases in performance were not adjusted for trends which were independent of this new control program. Because only one farm was studied, generalization of the results is very dangerous. The comparison in Table 1 is meant for illustrative purposes only. The values in Table 1 should not be considered statistically significant. b. Grunsell et a1. (1969) Grunsell and his colleagues (1969) took a different approach to determining the value of a preventive medicine scheme in England. Their sample consisted of 15 farms, not all of which were primarily dairy operations. The three-year project began with a visit to each farm by a farm management advisor. This was followed by a meeting on each farm of the farmer, his veterinarian, and an agricultural economist. After this introduction and initial appraisal, the farms received quarterly visits by veterinarians. The effectiveness of the scheme was rated according to the change in a number of performance indicators. These included 17 I yield per cow, milk sales per cow, concentrates per cow, margin over concentrates per cow, stocking rate, and margin over concentrates per acre. The farms were graded as showing "marked improvement" (7 farms), "some improvement" (5' farms), and "inconclusive" (3 farms). Grunsell was concerned with the benefits of such a program to both the farmer and veterinarian. He included some interesting discussion about the reactions of the 15 farmers and 10 veterinarians involved with the project. The ability of the advisors to suggest management improvements depended largely on the existence of good farm records. Overall success hinged on the farmer's organizational ability and willingness to accept management advice. This suggests that management ability may be the resource which varies most between farms. The grading procedure Grunsell used was so subjective it is difficult to argue with in detail. The main disadvantage of the approach is that it is impossible to generalize the results. In addition, performance trends that were independent of the program were not explicitly included in the analysis. c. Barfoot et al. (1971) Barfoot et a1. (1971) made an economic appraisal of a preventive medicine program for dairy cattle health management in Canada. They compared the performance of 27 herds visited monthly by veterinarians to a control group which received only emergency veterinary service (VS). The control farms were chosen to "closely resemble the organizational patterns and characteristics" of the farms participating in the preventive program. The period of study was two years. Five parameters related to herd health were monitored on all farms: milk production, days open, calf mortality, cow mortality, and culling rate due to health problems. In addition, the farms in the study were grouped according to expenditure on VS and drugs per cow. This cost ranged from $8.00/cow, for the group using only emergency service, to $35.00/cow, for the group with the highest "response" to the preventive program. 18 Probability density functions were determined for the five health-related parameters. These were programmed into a model with the intent to determine income per cow, over the cost of VS and drugs, given various milk and animal values. Their results showed that income over the cost of V5 and drugs was significantly greater for the farms spending $25, $30, and $35 per head than for a group of "similar" farms spending only $8 per head. This analysis does have the advantage of differentiating the health manage- ment program somewhat. In other words, by separating the herds into different expenditure groups, the authors were not really measuring the value of just one program. This was a step towards isolating the most profitable aspects of the preventive scheme. A control group was included which corrected for trends independent of the new veterinary program. This means that Barfoot et a1. did not just credit the program with all the improvements observed. The main disadvantage of this study was that the only cost measured was veterinary expense. Other costs incurred from increasing production were not included. Also, measuring health management by veterinary expenditure alone was misleading. The results imply that the more farmers spend on V5, the better off they will be. Low expenditure on VS may have been either a function of a farmer's ambivalence towards the value of veterinary care, or of the farmer's ability to administer health care independently. Finally, no mention was made of disease in Barfoot's work. A farmer's veterinary bill certainly depends on the degree of disease problems suffered by the herd. To measure the true value of a new program, some adjustment must be made for differences in disease prevalence. Perhaps this was corrected for in choice of a control group, but this was not explicitly stated. 19 d. Poterfield and Heider (1980) In contrast to Barfoot's findings that veterinary expense was directly related to profitability, Poterfield and Heider (1980) reported that large production gains could be achieved through preventive medicine programs that actually reduced veterinary expense per animal. Poterfield and Heider's survey consisted of 67 Ohio dairy farms that received regular visits from their veterinarians over an average of five years. The emphasis of the program was on reproductive and udder health. The average yearly increase in milk production for participating farms was 474 lbs./ cow compared to an annual gain of 265 lbs./cow for all Ohio cows on test. The average veterinary expense before the 67 farms received regular visits was $21.33 per animal compared to $20.13 per animal afterwards. Average total herd veterinary expenses rose, but this could be attributed to increase in herd size over the five-year period. The average herd size increased from 55 to 76 cows. Poterfield and Heider rightly compared the performance of the 67 herds on the program with their contemporaries. However, because they averaged all the results, there is no way of knowing which aspects of the regular programs were. more successful than the others. For example, the authors said that 38 herds received monthly veterinary visits, 16 herds were visited twice a month, and 13 were visited weekly. Which scheme proved most beneficial to which herds? In addition, Poterfield and Heider's study suffers from the by now familiar problem of not accounting for the non-veterinary costs necessary to increase milk production. These include feed, labor, and equipment costs. e. McCauley (1974) McCauley's approach did account for the additional costs needed to increase milk production. His data were from 117 Minnesota dairy farms over a period of two years. McCauley's goal was not to demonstrate the value of a preventive medicine program, but to measure the contribution of V5 in general to the income 20 of dairy farms. The farms in his sample primarily used emergency VS. McCauley's goal was accomplished through a production function analysis which had dairy enterprise income above feed cost as a function of' cow numbers, veterinary charges, drug expense, cows culled (for non-dairy purposes), and calves died. Cow numbers served as a proxy for all the capital and labor invested in the dairy enterprise. The calf mortality and cows culled figures were included to differen- tiate the severity of disease problems between herds. For 35 farms, McCauley also included a disease problem proxy which was based on mastitis incidence data. This did not prove very valuable. The number of cows culled for non-dairy purposes was found to be positively associated with profit. This is probably because the culling figure included those cows culled for low production. To be more meaningful as a disease problem proxy, it should encompass; only those animals sold because of a specific disease problem. Whenever VS reduces disease problems, it is contributing positively to gross income. With a severe disease outbreak, farm income may decrease even though VS expense goes up. Without correcting for the severity of the disease problem, VS could be seen to have a negative correlation with income, when the increased expenditure on VS actually reduced the amount of income lost. This is why some distinction must be made between the severity of disease problems on individual farms. McCauley found that an average increase in income over feed cost of $2.96 (6.55 in 1981 dollars) was associated with each dollar invested in veterinary service. Decreasing returns to size of veterinary expenditure were observed. Returns per dollar invested in VS were $8.03 (17.77 in 1981 dollars) for herds spending less than $6.00/cow (13.28 in 1981 dollars) compared to $1.82 (4.03 in 1981 dollars) for herds spending more than $12.00/ cow (26.56 in 1981 dollars) for V5. have 21 In conclusion, McCauley corrected for two deficiencies that he observed in previous studies. First, he corrected directly for feed cost and indirectly for all other production inputs by using cow numbers as a proxy. Second, he made some adjustment for differences in disease problems among herds. This latter correction is particularly important because he was dealing primarily with emergency V5, in which veterinary calls are more directly related to disease problems. However, McCauley's analysis would have been more powerful if more detail could have been provided about each herd's disease incidence. The advantages and disadvantages of these program evaluation studies are summarized in Table 2. Table 2. Some Advantages and Disadvantages of the Herd Health Program Evaluation Studies Hershler Grunsell Barfoot Poterfield et al. et al. et a1. dc Heider McCauley (1964) (1969) (1971) (1970) (1974) Account for Independent Trends? no no yes yes ? Differentiate Factors of VS Program? no implicitly somewhat no somewhat 9 Adjust for Additional Costs of Increased Production? no no no no somewhat Account for Severity of Disease Problems? no no no no somewhat 22 2. The Relationship Between Veterinary Service (VS), Disease, and Milk Production McCauley (1974) pointed out the need to adjust for the severity of each herd's disease problems in order to better estimate the value of VS. Over a broad range, we can expect additional investment in VS to decrease disease problems. It is also intuitively reasonable that the extent of disease problems influences expenditure on VS. This means that disease and VS influence one another, or: VS 6) Disease Previously mentioned mastitis and infertility research showed clearly that disease affects milk production. Veterinary service influences income (of which milk production is the major part) through disease, or: VS 6 Disease ‘9 Milk Production Evidence from Erb et a1. (1981) demonstrated a positive correlation between milk production and one disease. Erb and her colleagues found that cows with cystic follicles produced an average of 655 pounds more mature equivalent milk than non-cystic cows. Similarly, Shanks et a1. (1981) reported that the highest producing cows had the highest of selected health costs (drugs, veterinary costs, and some labor). These findings suggest the possibility that high milk production w more disease in some cases. It is intuitively reasonable that the increased stress of high production makes cows more susceptible to disease. If this is the case, then the relationship between VS, disease, and milk production can be expressed: vs 9 Disease 6) Milk Production There are two other possible explanations for the positive association between disease and milk production. The second explanation is that better managers recognize and treat more cases of disease. Actual disease incidence may not vary between herds, but the number of recognized cases might. Because better managers have higher producing cows, the correlation between production and 23 disease would be spurious in this instance. A third explanation is the farmer's tendency to tolerate more disease in high-producing cows. The extra income from high—producing cows makes it worthwhile to spend more for their maintenance. The total effect is probably a combination of the three factors. These are some examples of the complicated relationships that may exist between dairy cattle health management and profit. The various effects will have to be sorted out in order to adequately understand the influence of specific health management practices on profit. The strength of previous studies has been limited by available data. Computer-facilitated collection has made more comprehensive dairy cattle health data recently available. Several such computerized systems are described in the next section, along with other computer applications. E. Computer Applications to Dairy Cattle Health Management The data storage and processing capabilities of the computer are just beginning to be exploited for applications in dairy cattle health management. Kirk (1981) developed several routines for programmable calculators which aid in delineating the costs and benefits of various mastitis control procedures. The expected gain from mastitis control was some fraction of the estimated milk production lost based on California Mastitis Test scores. The expected cost of control was simply a tally of the costs of towels, teat dip, and antibiotic treatments proposed. The main limitation of Kirk's application is its reliance on CMT scores for estimating lost milk production. Because all cows are not tested using CMT on a regular basis, there is no consistent measure of mastitis prevalence. It may be argued that, because the costs of mastitis problems so greatly exceed the costs of prevention, there is little need to estimate mastitis losses more precisely than can be done using occasional CMT testing of a fraction of the herd, as Kirk suggests. Although it is obvious that mastitis can be very costly, some methods of controlling 24 or preventing mastitis are still controversial. This means that the benefits of some mastitis control techniques do not obviously far exceed their costs. More rigorous monitoring of mastitis prevalence in response to different treatments on individual herds would help dispel this controversy. Kirk's own research relating CMT scores to DHIA Somatic Cell Counts will facilitate careful mastitis monitoring. This research was discussed earlier. Kirk's technique for calculating the costs and benefits of mastitis control requires more detailed data to be powerful. The same is true for linear programming applications to dairy cattle health management. Carpenter and Howitt (1979) describe the use of linear programming (LP) for determining the most economical approach to the control of brucellosis. Linear programming is a mathematical formulation in which a series of linear equations are solved simulta- neously via computer. An objective function, such as minimizing the cost of brucellosis control, is solved given a number of constraints. Linear programming is only effective when the parameters of the objective function and constraints are clearly defined. For example, in Carpenter and Howitt's objective function, they included the cost of vaccination, market surveillance, personnel, and the value of cattle lost due to brucellosisgl Reliable estimates of these values must exist for their model to be of any use. For most diseases of dairy cattle, reliable cost data simply do not exist yet. The previously mentioned problems with making industry- wide estimates of mastitis and infertility losses serve witness to this fact. The above examples show that more data are required to fully utilize the analytical power of the computer for dairy cattle health management. 2’ All the elements of their tableau are not well explained. I have assumed that the large negative numbers in the objective function represent the value of cattle lost due to brucellosis. Another question arises about the constraint they have put on percent vaccination for 1976. As given in their tableau, it must be greater than or equal to 35 _a_n_d less than or equal to 25--for which no feasible solutions exist. 25 Coincidentally, other advantages of computers, namely their speed and ease of data storage and retrieval, are making collection of these data possible. The Dairy Herd Improvement program is primarily designed for recording milk production and for the genetic selection of cows (Crandall, 1975). Other computerized systems have been developed to improve herd's reproductive perfor- mance. Systems described by Britt and Ulberg (1970); Erb et a1. (1975); Gould (1975); Kelly and Holman (1975); Lineweaver and Spessard (1975); and Meek et al. (1975) are resigned to the retrospective analysis of reproductive perfor- mance. Cannon et al. (1978), however, describe a computerized herd health reporting system which is designed "to identify cows which show evidence of abnormal performance or are in high risk groups so that they can be examined and corrective procedures taken early." This goal is accomplished primarily through the provision of timely management reports. Separate computerized reporting systems are therefore available for milk production and reproductive herd health (including some non-reproductive disease reporting). The need to combine the two capabilities is recognized (Cannon et al.,. 1978). At best, however, this combination would still ignore farm infrastructure and labor and drug expense devoted to animal health care. A more comprehensive system, the Food Animal Health Resource Management System (FAHRMX), has been described earlier. The total data collecting ability of FAHRMX encompasses farm infrastructure, animal health histories, labor and drug expense for health care, genetic information, and milk production. All this is collected from individual commercial herds on a continuing basis. The balance of this report is spent discussing the potential application of these data in the cost-benefit analysis of dairy cattle health management. Chapter Three Data and Methods A. Introduction This chapter describes the sources of the new data available from the FAHRMX project, and potential applications of the data in cost-benefit analysis of dairy cattle health management. The first part of the chapter, Data Sources, contains: 1) A description of FAHRMX pilot herds using Dairy Herd Improvement Association (DHIA) indices. This serves the dual purpose of determining how representative the pilot herds are of all Michigan dairy farms, as well as demonstrating some of the limits of DHIA data. 2) A discussion of a questionnaire administered to some of the pilot herds, and its future uses. 3) A description of the content of the retrospectivefll data file entered onto mainframe computer. The second part of the chapter deals with uses of the new data. General concepts of cost-benefit analysis, centering on the partial budget, are introduced. Disease control expenses are itemized, as are different impacts of disease control. The problems with estimating these expenses and impacts are discussed. Several examples are used with these expenses and impacts in a partial budgeting framework. Finally, a statistical model is outlined which estimates one particular- ly evasive impact of disease, reduced lactating potential. Figure 1 shows the location of FAHRMX participants and sources of sample data, including the sources for both the questionnaire and disease data for the retrospective quantitative analysis. The purpose of the questionnaire was to obtain information about farm infrastructure and general disease control procedures. The ‘ retrospective quantitative analysis of disease and production records was a first attempt at estimating reduced lactating potential due to disease. fl“Retrospective" refers to the current study which relies on preFAHRMX data, while "prospective" indicates current or future FAHRMX capabilities. 26 Figure 1, Location of FAHRMX Participants and Sources of Sample / Participating arms which received questionnaire and had sufficient ‘ retrospetive disease data to be included in the quntitative analysis. A Participating . .s wh received questionnaire. D Other parti ipating farms. 1 O Participan without retrospective production records. I [I A [DO Frankenmuth A A ‘ Olonia ‘A‘ ADA D Qiichigan State ‘D University 27 28 Twenty-three farms are currently in the FAHRMX project. One of these had no retrospective DHIA records, which excluded it from Table 3. Twelve farms had had the questionnaire administered to them by the end of the summer of 1982. Because of administrative problems, no other farms have been surveyed subse- quently. Only eight of the twelve farms surveyed had good enough retrospective disease records to be included in the quantitative analysis of disease and production records. It should be emphasized that any data that were utilized in this study were available previous to the existence of the FAHRMX project's growing data banks. Such retrospective analysis pointed out deficiencies in pre-FAHRMX data, and modeling difficulties, that need to be overcome if FAHRMX is to achieve its goals. The retrospective study also provided baseline data by which FAHRMX's success can be measured. B. Data Sources 1. Description of Pilot Herds Four progressive Michigan veterinary practices were asked to select clients whose record keeping and management abilities could be augmented by a compu- terized decision support system. Therefore, the pilot herds are not a random sample of Michigan herds, but are probably typical of Michigan's better herds. All participating herds are visited at least once a month for reproductive exams and preventive health care administered by their veterinarians. All participants are also members of the Dairy Herd Improvement Association (DHIA), a service which currently provides production and management reports for 37 percent of Michigan's dairy farmers who milk 47 percent of Michigan's dairy cows. A comparison of selected characteristics of FAHRMX and Michigan DHIA herds is presented in Table 3. The FAHRMX herds were an average of 25 percent larger (91 versus 73 cows), and their milk production was an average of 8.0 percent higher (16,941 29 .338 No .382: com 83:33 omega ooucwmoBm .36 ms:— wctmuomc m3oo Ho 3253: emcee; on» E coEZc Eu... .3 Leo» con. coo-60.5 x52 1 3 6.280.. SID .3530:ch Eek—\m de J N D: me an m; 33.2 R MN (:5 :2 \wNon . q \waN \wQ: has \men \wa . _ \m_ am . 3 U»: R owm.o>< mz oN on. 3 an: N.N ”3 . D on on NN 3“; mm 8— ma «mm «N New.3 Ne w: _N mm”: 2 Nu me can n4 Rn.§ nu N2 8 IN; ON N: we own m.— 80.3 Nn mm m— mnN; N o: as mum m4 nsm:: me an ad m2 SN e: is 0R 3 2A.: 2 :_ 2 m2 wN wfi Ne nun dz woo. 2 ma 9: 3 m2 eN m2 we mum m2 ”3.2 we 3 3 dz 3 _N— R ”R m.N Z»;— ee 3 3 as; a me 3 can .3. 08.2 R .2 2 dz on no. «A. awn _._ New. 2 N: No N— S: R .2 c. in az an}: 3 n: : m2 2 mm 3 Nnn m4 tied an nn 2 mz 2 SN 3 8s «2 n S. 2 ea 3 m mz 2 am an mun Tn RN.3 . Ne .2 w nnw N NE 3 awn o4 Rm;— ma 3 N «Z “N o2 mm .mm mz Nan. 2 on :2 e ”N.— NN oN_ we com o._ {”8— 3 N2 n Re; 2 s: R is _.N 292 3 ”A e aNN; N N: as man n.N 23.2 ms _: n MZ mz m3 o: m2 m2 382 3 3 N In J on m2 3 can N .u 3n . 2 an 3 d A9 33 :30 in A933 coflaoocou 13.3.33 35:95 2500 Ev: Sou 3mm 930 £80 3235 can 1.352095 ow< 8e". :5 9.28 82:8 3 x52 .25 533. .HAowm: £05235 xoaun Op .52... .50» may cow <53..th _: 99.05 :3. 3). . 4) .5133...) i...) 11111111 -i if; - 30 versus 15,674 lbs./cow/year) in 1980. The next to the last row of Table 3 consists of weighted averages, with the exception of the first column. This means that the averages were adjusted for the number of cows contributing to the average. Some of the columns in Table 3 represent standard variables recorded for all herds on the DHIA program. Others, such as return over feed cost, are not required. This explains the missing values (NR5) in the last column. Most of the required features are reliable because they depend on, or are calculated from, data recorded by DHIA testers. These reliable variables include: number of cows, age, milk production, calving interval, days dry, and culling rate. There are problems with the days open and services/conception calculations because these rely on the farmers' recollection of breeding dates and not all farmers keep reliable breeding records. If only the latest breedings are reported, then the services/conception ratio will equal one. If no breedings are reported, then the DHIA computer takes the full length of lactation as the open period. This is probably why some of the days cpen variables are very large. DHIA presently does not separate those herds with complete reporting from those without. For their annual summaries, which are the source for the last line in Table 3, DHIA simply averages all the figures available from each herd, regardless of the fact that some herds have more complete information than the others (Thelan, 1982). This means that, for example, the number of herds contributing to the average of return over feed cost is less than the number contributing to the average of milk production. The lack of some performance indicators on some farms complicates interfarm comparison. FAHRMX is attempting to ensure complete recording of animal events data by providing more direct incentives, and by making reporting easier. It should be stressed that DHIA and FAHRMX have, and will continue to have, independent functions. FAHRMX will serve to augment DHIA by permitting the long-awaited marriage of health and milk production information. 31 B. Data Sources 2. Questionnaire a. Purpose and Description The questionnaire, which can be found in Appendix 3, had several purposes. There was a need to introduce FAHRMX to the farmers and obtain permission to utilize their disease and production records. The questionnaire helped depict the size, management practices, and livestock facilities of the farms. 1t documented the farmers' methods of dealing with common disease problems. The questionnaire also helped determine the costs associated with disease control that would not be apparent to FAHRMX either through the veterinarians' bills or the farmers' treatment reports. Examples of these latter costs include special facilities and equipment that the farmer uses for health care. The questionnaire also keyed in on routine treatment costs and times, such as that spent for dry cow therapy, so that the farmer need not report labor and drug costs for routine treatments. Finally, the questionnaire determined the quality of disease records kept before FAHRMX was utilized. The quality of these retrospective records determined whether the. farm could be included in the retrospective quantitative analysis. A modified version of the questionnaire will be administered on an annual basis to update existing information and ensure that farmers need only report daily events. Data from the questionnaire will be entered onto FAHRMX software, and therefore will be available for the analysis of the effect of different physical facilities or general management practices on herd health. b. Sample Size Limitations At the least, veterinarians can use this infrastructure data to aid in their hunches about the cause of certain disease problems. They can, for example, easily compare type of bedding to mastitis incidence on all the farms in the project. However, in order for such differences between herds to be statistically 32 significant, a specific number of herds must be participating with FAHRMX, depending on the variance of the parameter in question. The statistical criterion can be stated as follows: to be 100 (1-1) percent sure that the error x-n does not exceed "d" the required sample size is: n : [z‘lk 0/] 1 A This condition requires that something be known about the variance,d'2, of the parameter in question (Bhattacharyya and Johnson, 1977). There are two types of error relevant to statistical testing. The condition above requires that‘, or the probability of rejecting the null hypothesis when the null hypothesis is true, be specified. This is commonly referred to as "type 1 error." Type I error is usually considered the most serious. The null hypothesis is chosen so that the burden of proof falls on those who would consider rejecting it. The corollary to the null hypothesis in United States criminal law is "innocent." The alternative hypothesis is "guilty." 1n the correlation analyses that follow, the null hypothesis is that the parameters are not different than zero. Therefore, the parameter estimates will not be considered seriously unless the evidence against them being different than zero is overwhelming. Type 11 error is the probability of not rejecting the null hypothesis when the alternative hypothesis is true, which is usually represented by p (not to be confused with the parameter estimates). Given a fixed sample size, one type of error cannot be reduced without increasing the other. However, by increasing the sample size, both types of error can be reduced (Bhattacharyya and Johnson, 1977). The statistical model of milk production, which is described subsequently, succeeds because milk production varies between individual cows. Even though only a small number of herds are included in the sample, there are enough cows in those herds to make the model statistically significant. In the model, all the 33 variation between herds is assigned to one categorical variable per herd. The fact is that to detect statistically significant differences in factors that vary from herd to herd, instead of from animal to animal, many more herds are needed. B. Data Sources 3. Retrospective Data File The retrospective data file was developed to model the effect of disease on milk production. The resulting model is discussed in detail in a later section. The quality of pre—FAHRMX disease records was highly variable. For some diseases, it was unclear whether they did not appear on records because they were consciously not recorded, or because there were no cases of the disease over the period of study. On those farms that kept disease records, metritis and cystic ovaries were usually well recorded because they were diagnosed by veterinarians, and were recorded by the farmers along with the veterinarians' instructions for treatment. Of the 12 farms that have received the questionnaire, 8 recorded all cases of metritis and cystic ovaries (see Table 12). These were the diseases chosen for study simply because these were the only diseases for which data were available. The retrospective study was limited by the available disease data. Because FAHRMX is now building complete health histories of participating herds, this will not be a problem in future analysis. The indices of genetic milk producing potential--sire predicted difference (PD), dam index, and cow index--were obtained from DHIA semi-annual inventory reports. Sire PD is the expected extra milk production capability per year that a sire passes on to his daughter (when compared to a daughter of a bull with a PD of zero). Cow index is similar to a sire's PD in that it is a measure of a cow's ability to transmit milk producing ability to her offspring. The cow index depends on the individual's pedigree as well as her actual milk production. The dam index is simply a dam's cow index (ABS, 1975). 34 Genetic indices were unobtainable for many cows because few farmers kept all their inventory forms. This is another problem unique to the retrospective analysis because the genetic information can and should be one of the first lines of data entered on new animals in the FAHRMX project because it may be very useful in estimating potential milk production. Production data were obtained from DHIA monthly management reports filed at the DHIA office. These reports were photo-copied, and the pertinent data were copied by hand for entry onto mainframe computer. In future analyses, the transfer of DHIA production information will be done automatically via computer. The retrospective data file contains the following information for each cow: 10. ll. 12. 13. 14. 15. 16. 17. Herd Number DHIA Four-Digit Cow Identification Number Disease Code (0, for controls; 1, for cows reported as treated for metritis; 3, for cows reported as treated for cystic ovaries) Date of First Treatment for Disease Date of Onset of Lactation During Which Cow Was Recorded Sick Lactation Number Age at Calving (months) Dry-Off Date (end of lactation) Final Milk Production (pounds) Final Butterfat Production (pounds) Final Milk Production Adjusted to 305 Days Date of Next Calving Cull Code (Reason for Culling) Cow Index of Genetic Potential Dam Index of Genetic Potential Sire PD (Sire's Index of Genetic Potential) Total Days in Milk 35 18. Calving Interval (days) 19. Days in Milk at Date of First Treatment for Cystic Ovaries or Metritis 20. Season of Calving 21. Season of First Treatment for Cystic Ovaries or Metritis 22. Dollar Value of Production and other variables derived from the above as needed (see Appendix 4 for the complete data file). The file contains data roughly spanning the two-year period from 1979-1981. This span was chosen for several reasons. With a two-year span, the likelihood of obtaining matching production information for at least one calving interval was reasonably high. By just going two years back, the majority of the cows in the sample would still be in the herds. Finally, most farms that had pre-FAHRMX disease records had kept them reliably for about two years prior to receiving the questionnaire. This means that the retrospective data on most of the cows could be augmented by the prospective data currently being accumulated in FAHRMX data banks. The above information was entered for each animal beginning and completing at least one lactation within the range of complete retrospective production and disease records for each herd. This range of complete records was quite restrictive in some cases. For example, in a herd for which the span of complete records was less than one year, the number of cows having complete lactations within that period was only a small percentage of the total herd (see Table 15). In standard epidemiological jargon, diseased animals are called "cases" and non-diseased animals are called "controls." For the purposes of this study, case cows were those reported as having metritis or cystic ovaries during lactation. If the cow received more than one treatment for either of these two diseases during one lactation, only the date of first treatment was recorded. In the future, number 36 of treatments for disease will add important information concerning the severity of each case. If either a case or control were culled before drying-off, the reason for culling was noted (cull code), and the date of culling was entered as the dry-off date. The production data (final milk production, final butterfat, and 305-day adjusted milk production) at the date of culling were entered as the end-of- lactation figures, one difference being that the final 305-day adjusted production for culled cows was also mature equivalent adjusted. DHIA adjusts this variable so that farmers can judge the relative value of their culled cows. On Michigan DHIA management reports, mature equivalent production for other cows‘is only ex- pressed as a deviation from the average mature equivalent production of herd mates. If either a case or control were dryed-off but culled before calving again, no second calving date could be entered. C. Data Applications 1. General Concepts of Cost-Benefit Analysis a. Introduction At first, it may seem that organizing the costs and benefits of a disease control project and comparing them, such as is done in cost-benefit analysis, is a simple procedure. It is true that cost-benefit analysis would be greatly simplified if all costs and benefits were neatly timed, and if alternative resource uses need not be considered. However, it is a fact in animal production that the benefits of certain disease control procedures can accrue long after initial treatment. Like- wise, the costs due to inadequate health care can be far-reaching. Therefore, calculating the present economic value of future benefits due to today's treatment requires discounting the future benefits. Furthermore, unless resources are valued in comparison to their best alternative use, cost-benefit analysis will not arrive at the economic value of the proposed project (Gittinger, 1981). 37 Of course, cost-benefit analysis is impossible without sufficient empirical knowledge of major costs and benefits. Even with the detailed data FAHRMX is collecting, estimation of at least one significant category, reduced lactating potential due to disease, presents a considerable challenge. b. The Difference Between Financial and Economic Analysis Financial analysis deals strictly with cash income and cash expenses. Finan- cial accounting can be as straightforward as managing a checking account. Financial profit is simply the difference between money received and money paid out in a given period (depreciation and interest are usually deducted also). This calculation of profit can be thought of as the return to a farmer's unpaid labor and all other capital invested in the farm (Lipsey and Steiner, 1978). Economic analysis includes the opportunity cost of resources invested. Opportunity cost is the value of the resource if used in the next best alternative. For example, if a farmer has the option to work as many hours as possible in an off-farm job for $8.00/hour, then the opportunity cost of an hour spent working on the farm is at least $8.00. If a farmer's capital can earn at most 12 percent in an off-farm investment, then this is the opportunity cost of capital invested in the farm. In economic analysis, these opportunity cost values appear explicitly among costs. An economic profit of zero means that the farmer makes just enough money to be content with farming. However, a financial profit of zero means that the farmer is getting no return on "unpaid" labor and capital, or that the farmer is paying for the privilege of farming (Lipsey and Steiner, 1978). 2. A Basic Tool of Cost-Benefit Analysis: The Partial Budget The standard partial budgeting framework includes a list of factors which reduces profit and a list of factors which increases profit. Profit is reduced when costs are increased or income declines, and vice versa. Borrowing an example from Harsh et a1. (1981), assume that a farmer is considering increasing soybean acreage 38 by 40 acres and reducing corn acreage by 40 acres. The farmer expects $210/acre income from soybeans, compared to $262.50] acre from corn. But the soybeans cost less to grow. Cash expenses per acre are $54.66 for soybeans. In addition, soybeans require 4.7 hours labor/acre, which the farmer values at $4.25/hour. The corresponding corn expenses are $113.40/acre, and 4.1 hours labor/acre (also valued at $4.25] hour). The partial budget shapes up as follows. Partial Budget: Should the farmer grow 40 additional acres of soybeans and 40 less of corn? Step 1: Determine what increases profit of business. 1. Increased Income 1. Increased soybean income: $8,400.00 (40 acres * $210.00 income/ acre) 2. Reduced Costs 2. Reduced corn costs: (40 acres * $113.40 expenses/acre) 4,536.00 (40 acres * 4.7 hours labor/ acre * $4.25/ hour) 799.00 3. Subtotal $13,735.00 Step 2: Determine what decreases profit of business. 4. Reduced Income 4. Lost corn income: $10,500.00 (40 acres * $262.50 income/ acre) 5. Increased Costs 5. Additional soybean costs: (40 acres * $54.66 expenses/acre) 2,186.40 (40 acres * 4.1 hours labor/ acre * $4.25/ hour) 697.00 6. Subtotal $13,383.40 Step 3: Determine net change in profit (line 3 - 6) $351.60 It is clear from the above comparison that if the yield and price information used is reliable, then more money can be made if 40 acres of soybeans are grown instead of corn. However, the profit difference is small enough so that any risk involved in the shift might not make the shift worthwhile. 39 C. Data Applications 3. Itemizing the Costs and Benefits of Disease Control for Dairy Cattle How can partial budgeting be used to evaluate disease control for dairy cattle? The first step is to determine relevant cost and benefit categories. It is helpful to view disease control as reducing the impact of disease. Therefore, disease control is the cost and reduced impact the benefit. The cost of disease control should not be confused with the cost of disease. There have been many articles discussing costs of disease such as lost milk production due to mastitis. The issue here is by what degree does disease control reduce the "costs of disease." Table 4 identifies 13 factors divided into two categories: expenditure for disease control and disease impact. The text following Table4 discusses each item individually. Table 4. Itemization of the Costs and Benefits of Disease Control for Dairy Cattle Eigenditure for Disease Control l. Veterinarians' Service 2. Medicine 3. Farmers' Labor 4. Farmers' Special Health Care Facilities 5. Other Disease Impact 6. Milk Contaminated by Somatic Cells or Antibiotic Residue 7. Change in Feed Consumption 8. Reduced Feed Utilization 1n Youngstock 9. Reduced Lactating Potential 10. Death Loss 1 I. Culling 12. Lengthened Calving Intervals 13. Other 40 The items listed under "disease impact" generally represent losses of income due to disease. However, disease affects the herd by reducing some costs. Therefore, factors which both decrease income and decrease costs are included among "disease impact." Expenditure for Disease Control 1) Veterinarians' Service Veterinarians' service is an obvious cost of disease control. It is defined here as the cost of veterinarians' labor and advice. Medicine is included in a separate category to account for both that administered by veterinarians and farmers. Charges for veterinary service are assigned to individual animals by FAHRMX veterinarians at the time of treatment. 2) Medicine Each drug used in dairy practice has a code which is associated with a price per unit dosage in FAHRMX software. The cost of medicine, whether the veterinarians' or farmers', is automatically calculated by computer when farmers and veterinarians record treatments and dosages. 3 Gt 4) F armers' Labor and Special Health Care Facilities The extra labor requirements of sick animals is a commonly recognized cost, but until FAHRMX there has been no accounting for it on commercial farms. In general, little is known about the amount of health care which is administered by farmers acting alone. Animal health care expenses for farmers include the cost of their drugs, labor, and any health care facilities in which they have invested. The labor and facilities cost will be determined largely through data from the questionnaire mentioned earlier. The questionnaire can determine standard treat- ment times for farmer-treated diseases, as well as percentage use of special facilities (such as hoof-trimming tables) for each case of disease. Therefore, when most cases of disease are reported, the labor and facilities expense can be added 41 automatically. For cases of disease with uncommon treatment times, farmers are expected to record the amount of their labor spent, and special facilities used when applicable, for each case of disease. 5) Other The list provided is not intended to be comprehensive. There will most certainly be other expenses relevant to certain control programs. Impact of Disease 6) Milk Contaminated by Somatic Cells or Antibiotic Residue When codes for drugs which have milk withholding requirements are used, FAHRMX software automatically computes the number of pounds of milk withheld. Milk dumped because of high somatic cell count will also be recorded. Note should be made of alternative uses of contaminated milk, such as feeding to calves, and only the net loss considered. 7) Change in Feed Consumption Change in feed consumption may be positive, negative, or insignificant given different diseases. Because feed consumption can both increase and decrease due to disease, this category can represent both an increase or decrease in profit. The problem is academic, however, because FAHRMX herds presently have no way of recording individual feed consumption. Electronic identification of farm animals, combined with automatic feeding equipment, may soon provide individually con- trolled rations on many farms (Nott, 1982). Until then, an gp_r1_or_i decision must be made as to the relative importance of this category for each disease control procedure being analyzed. 8) Reduced Feed Utilization in Youngstock For cows, reduced feed utilization is measured in terms of reduced lactating potential. Weight gain might be an appr0priate performance measure for young- stock (non-cows). Holstein heifers commonly begin cycling at 600 pounds 42 regardless of age (Ax, 1981). If it is true that the average case of respiratory disease in calves causes a weight loss of 10—20 percent (AAPB Newsletter, 1979), then respiratory disease may cause a delay in getting heifers bred. Other diseases that inhibit weight gain should be charged with the delay in breeding that they cause. It is not likely that youngstock's feed intake will soon be monitored on a continuing basis on commercial farms. For this reason, the issue of reduced feed utilization due to disease in youngstock will probably have to be resolved on research farms. 9) Reduced Lactating Potential Each cow has an optimal productive capability, or lactating potential, that can be impaired by disease. If lactating potential can be reliably estimated, then the difference between healthy potential production and the actual production of a diseased animal can be charged to disease. Several factors complicate such a calculation: 1) Production potential changes with age (see #10 d: 11). 2) Disease can have long-term consequences, which may require lifetime disease and production information to detect. 3) Culling behavior as well as management practices influence the type of disease problems and the characteristics of the animals with the most disease problems. 4) There appears to be a joint influence between disease and milk produc- tion. All these problems will have to be dealt with in a model to estimate reduced lactating potential. The model is of such complexity that it will be considered in a section of its own. The problems of estimating reduced lactating potential are introduced here because some of them touch on later categories, such as death loss and culling, and lengthened calving intervals. In addition, with a reduction in lactating potential may come a reduction in certain costs as a result of having to handle less milk. These may include labor and equipment cost reductions. 43 10 d: 11) Death Loss and Culling Which animals die or are culled from the herd are determined by four factors: 1) Sale for dairy purposes; 2) Unpreventable circumstances such as natural disasters; 3) Preventable disease and accident problems; and 4) Selection by the farmer for genetic improvement of the herd. Here the focus is on factors (3) and (4). The interest is in how losses due to preventable disease and accident problems limit a farmer's culling choices based on production potential alone. To illustrate how disease affects culling behavior, let us first assume a culling rate of 25 percent per year regardless of whether disease problems exist or not. The culling rate seems to be primarily determined by the number of replacements that can be raised (see Appendix 5). A culling rate of 25 percent per year is representative of FAHRMX pilot herds (see Results). There- fore, in the absence of disease, it may be possible to cull all 25 percent based only on production potential. For example, consider a herd of 200 cows ranked by production poten- tial-S-l--cow #1 with the lowest potential and cow #200 with the highest--as in Figure 2. Then, without any disease problems severe enough in themselves to warrant culling, cows 1-50 may reasonably be culled. "" - s04! .. p4 'J‘l D 109 > cows culled due to Increasing Production Potential low production Figure 2. Culling Behavior in the Absence of Disease Problems Severe Enough to Warrant Culling 2/ Some problems with estimating production potential are discussed later. 44 Next, consider how disease would influence culling behavior. Data from FAHRMX pilot herds show that the probability of being culled due to disease problems is about 10 percent (see Appendix 6). These same data support the assumption that the chance of being culled is approximately random across production levels (see Appendix 7). Therefore, 10 percent of the 50 lowest producers, or 5 cows, and 10 percent of the 150 highest producers, or 15 cows, will be culled because of disease problems. However, the 5 of the 50 lowest producers would have been culled anyway. Disease has forced the culling of 15 cows that would have been kept in the herd had disease not existed. The net loss of replacement options due to the 10 percent probability of culling because of disease is therefore 10 percent of 75 percent, or 7.5 percent (10% '1 150 = 15 cows), not the full 10 percent. Figure 3 illustrates culling in the presence. of disease. Every tenth cow must be culled due to disease--for a total of 20. This just leaves 30 that can be culled because of low production. The 30 chosen are 1-9, 11-19, 21-29, and 31-33. Comparing the with and without disease scenarios, 15 cows that would not have been culled for low production have to be culled because of disease. In this example, these 15 are represented by cows 60, 70, 80, 200. JIJIIJIJ 100 :3’ 150 200 Increasin Production Potential 8 cows culled due to g low production 11111111 Figure 3. Culling in the Presence of Disease Problems Severe Enough to Warrant Culling: Assuming Every Tenth Cow Culled for Disease Reasons 45 Figure 4 illustrates how the loss of culling options lowers herd-producing potential. The assumptions on which Figure 4 are based are: l) The herd consists of 20 cows; 2) 25 percent, or 5, will be culled; 3) In the presence of disease, 10 percent, or 2, will be culled for disease reasons; and 4) In this example, these two cows are #10 and #20. Without disease, cows 1-5 would be culled. With disease, cows 10 and 20 must be culled, and 1-3 will be culled because they are the lowest three producers. Therefore, cows 10 and 20 are culled in the place of 4 and 5 when disease is present. The production potential lost is the sum of 10 and 20's potential minus the sum of 4 and 5's potential, which is represented by the shaded area in Figure 4. FAHRMX pilot herd owners are currently recording all reasons for death loss and culling. If a cow would not have been culled had she not been diseased, then her replacement cost should be considered as a cost of disease. But if she would have been culled because of low production anyway, then the disease should not be charged for her loss. Obviously, there is a need to be very precise about the reasons for removal. FAHRMX personnel should make this issue clear to farmers and ask them to ask themselves each time they report a cull, "Would I have culled her for low production anyway?" If the answer is "No," then her loss should be charged to the disease problem that caused her to be culled. Identifying cows culled strictly due to disease problems, and not low production potential, first requires a definition of production potential. The goal is a scheme illustrated in Figure 4 in which all cows in all herds are ranked by production potential independent of preventable disease. Only then can farmers say which cows they would have culled had disease not limited their choices. 46 Figure 4 Loss of Production Potential Caused By Loss of Culling Options 8 lost production potential i—> r d w '''''' F1 - —--u— 11— p — — I— —I- —II .— 4- d-— — F1 1 1 1 1 1 1 1 J 1 1 J 1 1 l J i 1 1 L l l I 1 If I l “T l ’l’ 1 *1 I I I *1 I I l 1 l 2 3 4 5 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Cow # 35’ increasing production potential a7 Estimation of production potential must adjust for the age of the cow. Younger cows' production should be mature equivalent adjusted. However, estimation of production potential independent of preventable disease is confound- ed by the positive correlation between age and disease. Because disease problems increase with age (Hlubik, 1979), it is debatable whether disease in older cows is always "preventableJé/ Increased age may make cows more susceptible to disease, in which case age causes disease to some degree. However, the positive correlation between age and disease also can be explained by the fact that older cows tend to be higher producers in which more disease is tolerated. The model of reduced milk production potential, which is discussed later in this chapter, must deal comprehensively with the interrelationships between disease, milk production, age, culling behavior, and other factors. Here we have hypothesized that the true costs of disease cannot be estimated until reasons for culling are better documented. But the fact is that the improved information provided by FAHRMX, although not perfected at first, will make culling decisions more rational. Once culling decisions are better understood, the information regarding disease costs can be further improved, which will further refine culling decisions. . . and so on with an iterative process. When the cows culled only for disease reasons, and not low production, can be identified, and when cows can be ranked by production potential independent of disease problems, then the loss of production potential caused by the loss of culling options can be easily calculated, as illustrated as in Figure 4. Other costs of replacement forced by preventable disease should also be considered. These costs include differences in breeding value not reflected in production potential. fi/ Prevention, as discussed previously, is an economic concept. Almost every disease or accident could be prevented if enough money were spent. This may also apply to age—related disease to some degree. However, age-related disease eventually leads to death regardless of expenditure. 48 Furthermore, there may be different cost reductions involved with culling different animals. For example, if the maintenance cost of high producers is more than low producers, then losses in production potential by the forced culling of a high producer would be somewhat offset by a reduction in maintenance costs. Costs may also be reduced because less milk is being produced. Finally, any differences in salvage value between culled cows should be considered. For example, if a diseased cow cannot be sold for meat, then her salvage value is going to be less than that of a cow culled strictly for low production. I 12) Lengthened Calving Intervals As research by Louca and Legates (1968) shows, a calving interval drawn out by disease (or any other cause) reduces milk production below optimum. Their research also supports the conclusion that gestation does not significantly affect milk production until after the first seven months of lactation. Therefore, major differences in milk production between the cow bred at the optimal moment and the cow with more days open would appear only after seven months. Figure 5 illustrates what must be the case if 12 months is an optimal calving interval (with a 60-day dry period). Any additional production because of more days in milk (Area B) must be offset by less production before the first dry-off date (Area A); otherwise the shorter CI is not Optimal. This picture of the effect of more days open on lactation seems to be contradicted by evidence from Oltenacu et a1. (1980) which supports the theory that delayed pregnancy frees more energy for milk production. Oltenacu and colleagues also provide evidence that there is a joint influence between milk production and days open, but they make no recommendations about optimal Cls for different production classes. Perhaps the association between high milk- producing cows and diseases which cause extended CIs is causing some of the confusion. The doubt raised by the Oltenacu data concerns the length of the 49 .fimENHQO on cu ho canoe NH new a woh< can» poumouu on umss < mou< m «09¢ < mon< :cwumuumq :w mzuco: ~u :ucca «a a sum: :00 05mm «mumuummuut MO :0 .m fiUn—VOHQ mummmmmmww- .\ n s” Ange messes m use waE cw mzucoa cmv mu :u:OE Nu a cum: zoo we newuuapoum Anew—V moummoq can «use: no: we AHUV ~m>houcu ucw>~mu umamumo mo coNuwcwmon m ousufim xams .n~ 5O optimal C1. The argument may continue over the shapes of the curves in Figure 5, but the definition of optimal CI will not change: additional production because of more days in milk must be offset by less production before the earlier dry-off date if the shorter CI is optimal}! Assuming that there is an optimal CI, and that some diseases cause reduced production potential by extending it, then this cost should be charged to disease. In order to do this, however, the optimal CI apart from disease must be defined. In addition, longer CIs cause fewer calves to be born per unit of time. For example, assume there is a herd with an average of 70 cows milking. If the average CI is reduced from 13 to 12 months, then 5 more calves will be born each year (70/12) * 12 - (70/13) * 12 = 5. The CI may be lengthened due to disease or management problems. The cost of sub-optimal calf production can be easily computed for those calves that would not have been chosen to replace culled cows. The loss from these animals is simply their salvage value. However, because diseases which lengthen Cls seem to be associated with high—producing cows, more offspring are going to be "lost" from high producers. This means that many calves unborn due to extended calving intervals would have been likely replacement choices. Because of lost replacements, production potential is lost. Estimating the cost of losing these calves is therefore similar to estimating the loss of culling options caused by disease. 13) Other The list provided is not intended to be comprehensive. There will most certainly be other impacts relevant to certain health management problems. Z, The only way that the longer CI could yield more milk in one lactation, and still be sub-optimal, is if it somehow causes a greater loss of milk in future lactations. 51 C. Data Applications 4. Examples of Partial Budgeting to Evaluate Disease Control Procedures a. Introduction The previous section pointed out many difficulties involved with estimating some of the items necessary to a partial budget evaluating disease control procedures. In some cases, a priori decisions must be made as to the relative importance of certain items because of the lack of empirical evidence about them. In this section, however, adequate knowledge of expenses and impacts is assumed so that the partial budget's application to disease control can be demonstrated. The values used in these examples are strictly imaginary. Because the numbers appear neatly in the partial budgets, and the difference is easily calculated, there is a danger of endowing them with too much power. It should be clear by now that many separate human decisions may go into estimating certain values. For this ever to be a useful decision-making tool, the underlying assumptions behind the more tenuous estimates must be explained to farmers. The variables in the examples represent the present values of the expected changes brought about by new disease control procedures. Present value calcula- tions are made so that present and future income and expenses can be compared. Consider the timing of income and expenses in the apple business. Substantial expenses must be incurred to plant an orchard, and several years must pass before any income is generated. Comparing the value of the expenses paid now against the non-discounted income received in the future would make the investment appear more profitable than it really is. Similarly, delays in benefits are expected from disease control procedures. For example, perhaps the major benefits from using udder wipes to control mastitis only appear several years after their initial use because production potential rises due to a decrease in culling forced by mastitis. Assuming that our statistical models can estimate when the benefits 52 come, and what they are worth in terms of milk production, then the present economic value of the benefits (and costs) must be calculated before they can be used in the partial budget (economic model). For an example of the importance of making present value computations, suppose two different potential investments exist, both with the same initial investment and undiscounted income returned over a four-year period, but with the income stream timed differently. Projectl and Project 2 both require a total investment of $750 paid immediately. Project 1 is expected to provide a steady income stream of $250 per year for four years. However, from Project 2, no return is expected until the fourth year, when $1,000 is expected. After four years, no other income is expected from either project. Table 5 compares the two projects at a discount rate of 12 percent. The discount rate is an estimate of the opportunity cost of capital, or the return that could be gained in the next best alternative investment. Although the two projects have identical undiscounted returns, because the timing of those returns is different, one is economically profitable while the other is not. If the discount rate accurately represents the Opportunity cost of capital, then any project with a net present value greater than zero is worthwhile. Therefore, Project 1 is viable, and Project 2 is not (Gittinger, 1981). 53 Table 5. Example of the Importance of Using Present Values Project 1 12% Net Incremental Discount Present Year Expenses Income Net Benefit Factor Value 0 $750 0 $-750 0 $—750 l 0 $ 250 $ 250 .893 $ 223 2 0 $ 250 $ 250 .797 $ 199 3 0 $ 250 $ 250 .712 $ 178 4 0 $ 250 $ 250 .636 $ 159 Total $750 $1,000 5 250 3.038 $ 9 Project 2 12% Net Incremental Discount Present Year Expenses Income Net Benefit Factor Value 0 $750 0 $ -750 0 $-750 l 0 O 0 .893 0 2 0 0 0 .797 0 3 0 0 0 .712 0 4 0 $1,000 $1,000 .636 $ 636 Total $750 $1,000 $ 250 3.038 $-114 54 b. Examples The first partial budgeting example considers using individual udder wipes as a preventive measure for mastitis. The only expenses are assumed to be the cost of the wipes and the farmer's labor. Let these figures represent the cost of using udder wipes for one year. The use of udder wipes for just one year may influence the herd over several years, by changing production potential through reduced culling because of mastitis, for example. In this example, the net present value of using udder wipes for one year is $40 ($200 - $160 a $40). Example 1: Use of Individual Udder Wipes to Control Mastitis Present Value of Expected Change Disease Control Expenses l) Veterinarian's Service 0 2) Medicine (udder wipes) $- 100 3) Farmer's Labor 5 -50 4) Farmer's Special Health Care Facilities 0 5) Other .__0 Subtotal $- 160 Disease Impact 6) Contaminated Milk 5 .10 7) Change in Feed Consumption 0 8) Reduced Feed Utilization in Youngstock O 9) Reduced Lactating Potential 5 +90 10) Death Loss 0 ll) Culling $+100 12) Lengthened Optimal Calving Interval 0 13) Other 0 Subtotal $+200 Total $ +40 The second example considers the feeding of colostrum to calves in the first 12 hours of life. The only expense is assumed to be labor. The example assumes that from previous experience colostrum feeding has strengthened calves so that their feed utilization goes up (item 8), their mortality rate drops (item 10), and 55 they eat more (item 7). The net present value of feeding colostrum for a set period is $250 in this herd ($300 - $50 = $250). Example 2: The Effect of Feeding Colostrum to Calves Within 12 Hours of Birth Present Value of Expected Charm Disease Control Expenses 1) Veterinarian's Service 0 2) Medicine (udder wipes) $ 0 3) Farmer's Labor $ -50 4) Farmer's Special Health Care Facilities 0 5) Other 0 Subtotal $ -50 Disease Impact 6) Contaminated Milk 5 0 7) Change in Feed Consumption $-100 8) Reduced Feed Utilization in Youngstock $+100 9) Reduced Lactating Potential 0 10) Death Loss $+300 1 1) Culling 0 12) Lengthened Optimal Calving Interval 0 13) Other 0 Subtotal $+300 Total $+250 C. Data Applications 5. A Model to Estimate Reduced Lactating Potential Due to Disease Much effort has been spent earlier in this paper pointing out the deficiencies of the. data used in other studies. The data used in this study also have many deficiencies, but that is because they rely on records which existed previous to FAHRMX. One of the purposes of this study is to direct FAHRMX to the information it should gather in order to meet its objectives. Although the retrospective data used in this study is deficient for some purposes, it differs from most other data sources in two important respects. First, the disease data are from commercial farms, instead of from research herds. Second, culled cows 56 are included in the sample. The importance of this second factor will become evident shortly. Another advantage of this study is that it has the prospective data capabilities to look forward to. Discussion of problems with the model can therefore take place at two levels. Some problems due to deficiencies in the retrospective data will be solved automatically by FAHRMX data, while others may be more lasting. The model to estimate reduced production, using retrospective data, is outlined after the significant problem of measuring production is discussed. No model can accurately estimate reduced production if that reduced production cannot be detected by current measurement schemes. a. Measuring Production Both the retrospective and prospective studies rely on DHIA for their production data. DHIA currently uses a monthly testing procedure which estimates the amount of milk produced by each cow. McDaniel et al. (1965) have calculated the correlation of DHIA's projections of 305-day milk production, made from one day's production measured monthly, to actual milk production measured at each milking. Early in lactation, the correlation is small but by the end of lactation (close to 305 days) the correlation is very high (.99). Therefore, DHIA estimates very well what a cow actually produces during her lactation by the 52d of her lactation. It was initially proposed that 305-day projections, made before the onset of disease, be compared to later projections or end-of-lactation figures in order to detect production losses due to disease. However, metritis and cystic ovaries usually occur early in the lactation period (see Results) when projection factors are unreliable. Assuming that the magnitude of peak production influences production throughout lactation, the projection factors for cows with diseases that exert their influence before the peak would not reveal lowered potential if they stay below potential over the whole lactation period. Figure 6 demonstrates this phenomenon. 57 comumNUem cw mama oom om -- )/ zoo ummmwmwc Co cowuuzuoca pmsuu< cocooaaeaa spa: Paaseoaaa .maucmuoa memueuoeu co snowman Co gummem aparmmoa o ohzuwm 0" ON x—wz nausea 58 In mid-lactation, DHIA averages monthly milk weights. If it is physiological- ly possible for a cow to recover suddenly from the stress of some disorders, then this practice of averaging monthly weights may miss "dips" in production between tests. Figure 7 shows the extreme case where a cow is stressed and recovers to her potential between DHIA test days. In this case, the monthly testing procedure would not detect any milk loss. However, DHIA's procedure will also overestimate milk lost in other instances (see Figure 8). It is also possible that when a disease occurs in mid-lactation, the cow never completely recovers from the stress. In this instance, a difference in projection factors should be detectable. Figure 9 illustrates the case where a mid-lactation disease causes a detectable change in projection factors. At the fifth test day, the projected production can be represented by Area A. Assume that the cow is stressed immediately following the fifth test day; then, at the sixth test day, projected production could be represented by Area A minus Area B. A comparison of projections at t 5 and t 6 would show a net loss of Area B. The advantage of using changes in projection factors is that cows can be used as their own controls. A disadvantage of projection factors is that they have not been determined independent of disease. Therefore, their use would probably underestimate the actual reduction due to disease. The fact that the retrospective data included only diseases which occur early in lactation (metritis and cystic ovaries) precluded the use of projection factors in the retrospective analysis. The only option left was to use end-of-lactation production figures. These are the most reliable and will account for production "dips" as well as any monthly testing scheme can. The use of daily milk weights would circumvent many of the problems mentioned here. A few FAHRMX participants currently have the equipment to measure daily milk weights. Research is being done to use deviations in daily 59 :omumuueg cw mxeo Nae ammo eox.n "on Nee use» geese "ma cur—:30 zoo emmemmwo ea cowuuavoca Peauu<\ / \ cowuuzpoga xpp: pmwucmuom / , ~— ’ spa: nausea mousvoooum mcwumoh xflgucoz x2 wouoouopca cofluuavone xfiflzirw :mwo: n enawwm xmp «mow zuxmm cu ll m be. :8 eta :cwumuomq cw mum: use umou cupaom ea in a O 60 my ouspouona .w. wcmumow xmzucoz xm my vouwemumm acmuusvoum newuuapoue gm“: sou powwow“: we comuuztcpe Hmsuu< compospcpc umc~ mo ccmumewumoho>c i ouscooche mcwumoh >~xuccz a: poucemumopo>o newuosc0pc xuw: cw :QNQ: m ohaumr 61 Nee anon aux.n no» eoaaauoee em mama Nee use» euc.c um» mom . as me o _ _ _ :m: me< ..<.. mmL< § Hmfiucouoa ucmumuumq wouzvom unease on mucuumm cofiuuoHOHQ zen mom mo om: x—vz mveaoa w ousuuu 62 production as a diagnostic tool (Anderson, 1982). Combining daily milk weights with health histories on FAHRMX software would facilitate this research as well as remove doubts about how much milk a cow actually produced. b. Sc0pe of the Ideal Model In a previous section, some problems were introduced that will have to be dealt with in a model to estimate reduced lactating potential caused by disease: 1) Production potential changes with age. 2) Disease can have long-term consequences, which may require lifetime disease and production information to detect. 3) Culling behavior as well as management practices influence the type of disease problems and the characteristics of the animals with the most disease problems. 4) There appears to be a joint influence between disease and milk produc- tion. Some accounting of factors (1) and (3) can be made using retrospective data. Factor (2) will be accommodated automatically by the accumulation of FAHRMX health histories combined with DHIA production information. However, factor (4) presents difficulties that will not be solved so easily. First, the format and capabilities of the model using retrospective data will be discussed. Then, some possibilities for future improvement will be outlined. c. Format of the Statistical Model Using Retrospective Data Both age and season of calving have been shown to significantly influence milk production (Miller et al., 1970). The variation between herds is also large enough that DHIA will soon make adjustments for herd effects on production (Thelan, 1982). The length of lactation has an obvious influence on production as well. Finally, disease has been shown to affect milk production (Erb et al., 1981), and exploring this effect further is one of our goals. Information concerning all these factors is included in the retrospective data file. 63 If it assumed that age, season of calving, herd, days in milk, and disease combine to determine milk production, then single-equation multiple-correlation analysis can be used to determine their various contributions. To estimate reduced lactating potential, the goal is to attach a cost, in terms of milk production, to each case of a particular disease. Although correlation analysis cannot prove causality, it can be used to show the strength of association between two or more phenomena. With milk production chosen as the dependent variable, and age, season of calving, herd, days in milk, and disease chosen as explanatory variables, correlation analysis will estimate the (population) mean value of milk production in terms of the explanatory variables. If the model is specified correctly, then the disease parameter estimates (B's) will estimate the average change in production caused by each case of disease. The expectation is that disease parameter estimates must be negative in order to adequately represent the detrimental effects of disease (Gujarati, 1978). The method of estimation used was ordinary least square (OLS). To obtain unbiased estimators using OLS, six assumptions must be valid: Assumption 1: The conditional mean value of the population disturbance term u., conditional upon the given values of the explana- tory val'iables (the X's), is zero. Assumption 2: The conditional variance of U1 is constant, or homoscedastic. Assumption 3: There is no autocorrelation in the disturbances. Assumption 4: The explanatory variables are either nonstochastic (i.e., fixed in repeated sampling or, if stochastic, distributed independently of the disturbances, ui. Assumption 5: There is no multicollinearity among the explanatory vari- ables, the X's. Assumption 6: The u's are normally distributed with mean and variance given by Assumptions (1) and (2) (Gujarati, 1978). With milk production as the dependent variable, and age, season of calving, herd, days in milk, and disease as explanatory variables, the equation looks like: 64 Pounds Milk Production = so + 8 Age at + Calving 82 (0,1 Variable for Spring Calving) + 83 (0,1 Variable for Summer Calving) + 84 (0,1 Variable for Fall Calving) + 85 (0,1 Variable for Herd 2) + 86 (0,1 Variable for Herd 3) + . . . + 8“ (0,1 Variable for Herd 8) + 812 (Total Days in Milk) + 813 (Total Days in Milk)2 + 814 (Total Days in Milk)3 + 815 (0,1 Variable for Metritis) + 816 (0,1 Variable for Cystic Ovaries) + unexplained error Milk production was the total milk produced per lactation. Other researchers adjusted production to a standard 305 days. We felt some information might be lost by following this example. Time is accounted for in our model by the days in milk terms. The season of calving, herd, and disease variables are all categorical variables, i.e., they can only take on the values zero or one. One less than the total number of each group of categorical variables is explicitly included in the model. For example, there are four seasons, but only three are explicitly included in the model. Production of a cow calving in winter, in herd 1, without either metritis or cystic ovaries, is estimated as a function of the intercept term, the age term, the days in milk terms, and the unexplained error. As its name implies, the error term accounts for all the variance not already "explained" by the explanatory variables. 65 In a model of similar specification (i.e., single-equation correlation analysis) Erb et al. (1981) found positive disease parameter estimates for cows with cystic ovaries. Cystic cows were found to produce an average of 655 pounds more mature equivalent milk than non-cystic cows. Using our previous interpretation of the model, this suggests that each case of cystic ovaries is "worth" about 655 pounds of milk, which could erroneously lead to the recommendation that farmers welcome the disease. There are obviously other factors as yet "hidden" from this model. What Erb's evidence shows is a positive association between cystic ovaries and cows which produce a lot of milk. How else could this association be explained? One strong possibility is that farmers tolerate more disease in their high producing cows. As an example, suppose a herd is categorized into low, medium, and high producing groups as in Figure 10. Assume that one-third of the cows are culled each year. Also, suppose that cows can be categorized into those with no, modest, and significant disease problems. If the farmer were to follow a culling procedure that ignored disease, equal numbers of cows would be culled from each group as depicted in Figure 10a. Therefore, if an investigator were to sort cows into the three disease categories and examine the impact on milk production, an unbiased estimate of the impact would be obtained. In contrast, suppose the farmer takes disease into account in culling decisions, and that a higher lactating potential is required to retain a cow that has disease problems. The results would be similar to Figure 10b. Therefore, if cows were sorted into the three disease categories, milk production for the cows remaining in the herd would be seen to increase as disease increases. Biased estimators of the impact of disease on milk production would result because of the confounding effect of the farmer's culling procedure. Erb's analysis effectively excluded culled cows because for her "regression analyses" she chose a subset of 810 animals with "complete records," a complete FigurelO Illustration of Culling Bias Significant Some None Significant 66 O. O .0 -_---..A-.-- Shaded Region Depicts Culled Cows I Some None 0 IOOIOOOOIOOOIOO 0.0.0.0....O... I .:...:.0.0.0.0...0...I.0.0.0.. O I 00.1, no 0... II cones-00...... c.0000... cocoon-000.00. 0.00000 .0. 0.......I....‘I‘.‘IOOOIOO Production Potential A gure 10b IF! [-1. Figure 103 Disease 'dering Sl Cows Culled Con Cows Culled Ignoring Disease 67 record being defined as the middle of three successive lactation periods. The inclusion of culled cows in our sample should help correct the bias by "saving" disease information about culled cows. We would expect that inclusion of culled cows will lower the cystic ovary parameter estimate (make it more negative). Although culled cows were included in our sample, we cannot be certain that the parameter estimate for cystic ovaries will be negative (representing its "true" value). This is because there are probably other significant factors creating the association between high milk production and disease for which our model does not yet account. A further explanation may be that better managers recognize and treat more cases of disease. Actual disease incidence may not vary between herds, but the number of recognized cases might. Because better managers have higher producing cows, the correlation between production and disease would be spurious in this instance. This "reporting bias" is a "bug" that will be very difficult to remove. A third possibility is that the stress of producing a lot of milk actually makes cows more susceptible to disease. If this is the case, then disease and milk production are jointly determined to some degree,-8-/ and a single-equation model of milk production including disease is not legitimate because OLS assumption number four is contradicted. Rather, a set of simultaneous equations must be considered. Such a system of equations exceeds the limitations of the retrospective data. The simultaneous equations problem is discussed in the next section along with the prospective data probably required to solve it. -8-/ The vector linking the joint determination of days open and milk produc- tion, discovered by Oltenacu et al. (1980), may well be cystic ovaries. That is, cystic ovaries are jointly determined with milk production. Because cystic ovaries lengthen the open period, it may only appgar as if milk production has a direct effect on the open period. 68 d. Possibilities for Future Improvement of the Model: Simultaneous Equa- tions In single-equation regression models, such as the one discussed previously, the dependent variable is listed as a function (linear in the parameters) of two or more explanatory variables. The fundamental assumption of this procedure is that causal relationships, if they exist, move only from the explanatory variables to the dependent variable. If the stress of producing large amounts of milk does make cows more susceptible to disease, then a single-equation model is no longer appropriate because disease and milk production are jointly determined. Here two different effects of disease must be distinguished: (1) disease as a result of poor management and (2) disease as an "inevitable" result of milking cows at maximum capacity. This latter problem involves developing a set of equations to be solved simultaneously. Another way to state the simultaneity problem is that if disease is correlated with the error term in a single-equation model of milk production, then there is no way to assess the separate influence of disease and the error term on milk production. This is a violation of OLS assumption number four (Gujarati, 1978). The equations discussed previously had milk production as a function of age, season of calving, herd, days in milk, and two diseases, or: Milk Production* = f (Age, Season of Calving, Herd, Days in Milk, Disease 1*, Disease 2* + error) where the asterisks (*) indicate jointly determined or endogenous variables. Each endogenous variable requires an equation which uniquely determines it (Gujarati, 1978). With three endogenous variables, three equations are needed such as: Milk Production* : f 1 (Age, Season of Calving, Herd, Days in Milk, Disease 1*, Disease 2*, Genetic Potential + error") 69 Disease 1* = f2 (Milk Production, Age, Season of Calving, Herd, X + error'") 1 Disease 2* = f3 (Milk Production*, Age, Season of Calving, Herd, X2 + error'"') where X1 and X2 are exogenous variables yet to be determined. The unique determination of the equations is called identification. To identify the two disease equations, two exogenous, or predetermined, variables are needed: X 1’ which is correlated with Disease 1 in individual animals but not milk production, and X2, which is correlated with Disease 2 in individual animals but not milk production (Gujarati, 1978). It is proposed to use lagged dependent variables for X l and X2. Coleman's (1982) research on the recurrence of disease problems should be helpful in this regard. Because FAHRMX is currently amassing disease histories, it is possible that a lagged disease variable could be used in the prospective analysis. Both the rank and order conditions must be met for identification (Gujarati, 1978). This segment on the model to estimate reduced lactating potential due to disease began with criticisms of monthly estimates of production. It concludes with a warning about what can be measured even with perfect knowledge of production. e. A Point of Clarification: Measuring Reduced Lactating Potential in Animals Treated for Disease In a strict sense, the foregone production due to lack of treatment for specific diseases in commercial herds cannot be measured. When a cow is diagnosed as ill, the animal is either treated, culled, or left untreated. Presuma- bly, the decision to keep an untreated sick cow is based on the assumption that the disease is not very detrimental. This leaves a sample devoid of animals untreated 70 for serious diseases. For those animals that either die or are culled due to disease, foregone production can be estimated as described earlier. Because veterinary treatment has already been justified for the surviving animals, the problem remains to calculate the value of treatment in cows that have already been treated. This is not a fair measure of the value of treatment. In fact, treatment may be very effective in preventing detrimental effects of disease, that is its purpose. What can be measured is the value of changes in treatment procedure. f. Conclusion to Methods To this point, it should be clear how the costs and benefits of disease control for dairy cattle can be compared once they are enumerated. A proposed model for enumerating lost production potential has been presented; however, difficulties with the model's specification are expected. The next chapter tests the model with data from FAHRMX herds that were available previous to FAHRMX. Chapter 4, therefore, contains much discussion about the representativeness of the sample data. Chapter Four Results A. Introduction This chapter is divided into five sections. The first lays out some results from the questionnaire which depict farm infrastructure and management factors pertaining to animal health care on the 12 farms surveyed. The second compares the sample disease data from 8 of these 12 herds to that of other studies. The third compares the characteristics of culled cows in the sample of eight farms to those in other culling studies. The fourth compares the results of the correlation analysis with and without culled cows. And, the fifth presents the results of a correlation analysis including genetic indices. B. Questionnaire Results The herd size of 22 FAHRMX participants was given in Table 3. Table 6 shows the size distribution of the l_2 surveyed farms. Table 6. Herd Size, 12 of 23 FAHRMX Participants, 1980 No. of Cows < 50 51-75 76-125 126-157 No. of Farms 3 l 6 2 Source: DHIA Annual Herd Summaries, 1980. Table 7 shows the milk production classes represented by the 12 farms. 71 72 Table 7. Milk Production Class, 12 of 23 FAHRMX Participants, 1980 Pounds Milk/Cow/Year No. Farms 15,000 - 15,999 3 16,000 - 16,999 2 17,000 - 17,999 5 18,000 - 18,999 0 19,000 - 19,999 2 Source: DHIA Annual Herd Summaries for Holsteins, 1980. Table 8 depicts the housing of lactating cows on the sample farms. Table 8. Housing for Lactating Cows, 12 of 23 FAHRMX Participants, 1982 Tie Stall, Free Free Stall Tie Stall Stall, and Pasture No. of Farms 9 2 1 Table 9 shows the location of milking on the 12 farms. As might be expected, the nine farms with milking parlors coincide with the nine farms with free stalls in Table 8. Table 9. Location of Milking, 12 of 23 FAHRMX Participants, 1982 Milking Parlor Stalls No. of Farms 9 3 73 Table 10 shows how dry cows are housed on the sample farms. Three of the farms with free stalls for their lactating cows only provide loose housing for their dry cows (see Table 8). Table 10. Housing for Dry Cows, 12 of 23 FAHRMX Participants, 1982 Free Stall-s Loose No. of Farms 6 6 The existence of special health care facilities should also be noted so that their influence on farm profitability or disease incidence may be studied. We received varied answers to the question, "Do you have a facility where you isolate or give special care to sick animals?" Many box stalls had multiple uses, some of which were not related to herd health. Some farms had nothing that could be classified as an isolation facility. Others had relatively elaborate facilities including hoof trimming tables, squeeze pens, catch pens, and head gates. Table 11 shows the location of calving on the 12 farms. Table 11. Location of Calving, 12 of 23 FAHRMX Participants, 1982 Maternity Own Mostly Stalls Stalls Outside No. of Farms 10 l l The low calf mortality reported by the 12 farmers, zero to 5 percent for females by the time of weaning, draws interest to calf care on the sample farms. The farms raise all their replacement animals. Age at weaning ranges from 4 to 12 74 weeks with an average of 7. Six farms have calf barns. Only one of these has a heated nursery. Five farms use calf hutches. Table 12 addresses the degree of disease reporting on the sample farms before FAHRMX was used. For some diseases, it was unclear whether they did not appear on records because they were consciously not recorded, or because there were no cases of the disease over the period of study. The least common diseases among cows included pneumonia, hardware disease, diarrhea, pink eye, and bloat. The records consisted of date and treatment. They were either on MSU issue individual cow cards or cow folders, or were merely kept in a notebook. Cystic ovaries and metritis have the advantage of being primarily veterinarian diagnosed. This allows for consistency in diagnosis as well as reliable reporting. The common practice on these farms is for farmers to record the veterinarians' diagnoses at the time they make them. On the basis of these records, 8 of the 12 farms were chosen for the quantitative analysis of disease and production records. Tables 13 and 14 reveal the degree of dependence of the farmers on their veterinarians. It is important to note the amount of health care that farmers perform themselves in order to calculate the total value of health care received by the herds. The degree of veterinary self-sufficiency may be an indicator of management prowess, or simply a function of proximity to the veterinarian's office. Tables 13 and 14 are based on the farmers' judgement of the percentage of self-treating that they do. The prospective data base includes empirical evidence of the amount of self-treating actually done. All but one of the twelve sample farms dry treat all their cows for mastitis. Ten teat dip regularly. Two do not teat dip regularly. Five use the California Mastitis Test or the DHIA Somatic Cell Counting service. Seven farmers use only clinical signs to diagnose mastitis. 75 Table 12. Extent of Disease Recording Before the Utilization of FAHRMX, 12 of 23 FAHRMX Participants, 1979-1981 Number of Farms That Recorded All Cases Fewer .— O Cystic Ovaries Metritis Displaced Abomasum Milk Fever Hardware Disease \IV‘JU-PN Ketosis Bloat Mastitis Pink Eye Pneumonia 10 11 ll 11 12 Lameness Or—t—v-t—NVIVIVoVOO Diarrhea Table 13. Milk Fever, Percentage of Farmer Treatment, 12 of 23 FAHRMX Participants, 1982 Percentage of Cases Self-Treated None 90% or More No. of Farms 6 6 Table 14. Retained Placenta, Percentage of Farmer Treatment, 12 of 23 FAHRMX Participants, 1982 Percentage of Cases Self-Treated None 90% or More No. of Farms 4 8 76 C. Comparison of Sample Disease Data to That of Other Studies Figures 11 and 12 show the distribution of days in milk at first treatment for metritis and cystic ovaries. The histograms support Shanks et al. (1981) conclusion that most of the treatment expense occurs in early lactation. In Figure 13, the incidence data for metritis and cystic ovaries have been combined in order to compare them with Shanks' et al. (1981) and Hansen's et al. (1979) reproductive disorders cost curves. Shanks included treatment costs for metritis, pyrometra, discharges, adhesions, cysts, retained placentas, tears in the reproductive tract, difficult calvings, as well as postpartum and other reproductive exams. Hansen excluded palpation labor and expense, but otherwise included the same disorders among reproductive costs. Shanks' data are from two research herds and cover about 1,000 lactations. Hansen's are from only one research herd but include about 2,500 lactations. It appears as if Hansen was more rigorous about recording the amount of farmer labor spent for animal health care. This might account for the difference between Hansen's and Shanks' reproductive cost curves. From these comparisons, the importance of disease in the early stages of lactation is reaffirmed. The coincidence of total treatment expense and frequency of first treatment is probably not very surprising. The similarity across herds, however, is remarkable. It is important to stress that the costs tabulated by Shanks and Hansen were treatment costs. They omitted the major impact category of reduced milk producing potential. D. Characteristics of Culled Cows The inclusion of culled cows in the sample correlation analysis is expected to have an important effect. Therefore, it is crucial that culled cows included in the sample are representative of the whole population of culls. For some herds, the span of complete production and disease records was quite short. It was thought that particularly short spans would bias the sample in favor of culled cows because 77 «coaumopk umpmm mo moewh um x_w: Cw mum: eee cos em a p - — — 11— . ,_ H1 _ :4 . mete: oo_ae seems .nssseSoz new «cosumouh «meme mo oE_b um x_wr :« mxmo NH onsuwm 414 l Na 0“ ow 9N wN Nm on an Aouanbaag acosumohh smoke we came um xng :N mxmo Aouanbaig ccm Gm~ cow cm A _ _ _ — — — d u a II V all 0 ll w 1Te~ mete: oosaa cease .nosne>o unease sou useEuaouh umuwm we came an xmwz cm mxma Na assume cowueuoep cw mzea omm com CNN ovN ofiN om“ omH ONH co so on i. _ _ _ _ _ d 1 . .i d1 d m JI- 4 16m. 0 . S 11 I DJ 184 .I- 3 1T 18.“ 38 8:26.85 ' om 1T P3258 .2255 1oo.N O O 4 N. mm 1:1 'J IQm.N mvuwcumz pee $238 miuuzwwmwmm 3:25 B a 186 mm_ca>o uwumxu cm . . so mm 1T :8 .86 muemuwuc— . 323.5 miuuavocomc Pcomcm: I 354:8 ca 11 18% m. .11. vem.v 1T 4 39m 4 mean: oosna «amen saga mwuwuuoz can mowhm>o owumxu you name oucopfiocm pocwnecu .mcwaflau o» :oflusnwuumwa umou popuomflo o>fiuozp0pnoz mxemzm one m.:om:m: mo cemwhmaeou m_ assume weep—co 80 their lactation periods tend to be shorter. However, the data presented in Table 15 dispel this fear. Although the sample culling percentages vary significantly from the percentages calculated by DHIA in 1980, the weighted mean culling percent- ages are both 24 percent (the last row in Table 15). The information in Table 16 gives further evidence to the representativeness of the sample culls. Table 16 compares the reasons for culling in other samples to the data from eight pilot FAHRMX herds (last column in Table 16). Although the culling percentages in the Cummins' data are not necessarily annual figures, and some of the classifications of culls differ between studies, the similarity between the percentages from the various studies is evident. The data from which the last column in Table 16 was derived are contained in Table 17. In Table 17, the mean mature equivalent production of those cows culled for dairy purposes is expectedly high. The magnitude of the production potential lost because of culling due to disease can be seen in the high mature equivalent production of cows culled for the following reasons: physical injury, mastitis, sterility, milk fever, illness, and leg problems. The low mean days in milk figures for those cows culled because of udder problems (89 DIM) suggests that the most severe udder problems occur early in lactation. E. Correlation Analyses Compared: Sample With and Without Culled Cows The results of the first regression excluding culled cows are contained in Table 18. The R-square value of .63 means that only 63 percent of the total variation in milk production is explained by the model. Given that all the OLS assumptions hold, it is evident that several parameter estimates are significant. The significance level represents the probability of rejecting the hypothesis that the true parameter value is zero when it is in fact equal to zero. Therefore, the smaller the significance level, the smaller the chances of falsely assuming that the parameters are different than zero. The parameter estimates represent the 81 n oases aonae a a "is .nano; ”Ha ea nsoo . snoop \ A one: naoo a . .eno; asao «U u w u mpuouoa ouomesou mo mueo> o fiasco Hench\m-=u gauchvp .oHanwm>u one: mvuooou ccwuuspoum was omaomwv o>wuooAmouuon anon sown: avenues mouav och“ eeN oeN can: eoueunoz Hw\nN\n 0N mm Cu Ne an.” m.~N t m>\N\oH m ~w\mN\m NH mm n ma new. m.o~ i om\e\H~ n Hm\nN\o 0N mm ma mm NH.“ o.v~ tow\ea\v o Hw\c~\o wN 0N we wag ow.” m.n~ i ow\n\~ m Hm\MN\m VN nv mg a“ Ne.“ o.n~ i ow\mN\n v Nw\~N\H ~n m" mm ea“ oo.~ o.oN t ow\nN\m n Hm\o~\o OH mN NN ~m Ne.“ o.N~ i om\mm\~ N Hw\~\o an N~ n" no eo.~ c.0N t aN\NN\m a <~:o ou nuao> use mafiau "such nsou «such Amheo>v :emw anzucozv cemm ampuooom one: ucfipuouu< one" ponmsu ucoonoa mo guano; mo caucoa ouoHnEou CH pee» won me :aaw voumau ucouuom enmeam o>fiuoomnouuom xzuzou meanaoh 030m «nouuaun amouua enouucz as: one: «camaozm .vouounuuou no: apnea a. awn "voweunuuou one none: w“ vo_~:u «mN _ «m.ee ace. www.9a mace" see" pc.aa .eaoh A.” e.~ o.o “.4 6.6 a." ou< an" v.0 nN.N~ oN vs.v N.v xy:-:~ Ac" n.a aa.n e.n e.» an." onaonme .m— m m.~ o.~ oeoesoo< as" ~.~ 0.4 oe.~ ~.v~ ".6. case an“ n.« m. h.e eoueonea .Nu m ou.a n.~ o.o~ - o.a soeoo .~_ m.o v.~ 6.. m.~ Nuausa< x~_z no" m.o o.N . o.~ sou>eeam as o.~ ease as ~.n o.a a.“ oaoa eee nae; .5 o.N c.m ~o.m n.o n«u_un=z Ac a.~ ~_ u-.n. 'o.a_ o.u_ ~.e teas: an e.e_ ~._~ me.e N... se.ee .e o.en N.oq cw Na.on N.mN o.o~ ~.NN .eoue :6; An ae.a n._~ oo.m_ _.o~ Noassnoum an on c.m~ o.vN eouuuavouaoz a" N " cannon .eN . eo>au «oz ae.en eo>so ooz eos.u “oz «Nu «emu «causau we can: ~e3==< nesoonioz neg neaoun~oz ana.a voa._ “an.“ ae-.n Non.n nsou . eea_-aaa. ~aaa-.ma. aoa~-noa_ noa_.ema~ ama_.ama~ Noa— kaaom no ooae seaweed: engage“: qweuou_-u a~ea>~xneeom avauonm need—hogan: «no» no: enquauea «Cu-BIG Sing :0 «Em muonouz hozuoon xuo~> 93> toga—.2 a o_onae a swag: a e~oen< . sauce—«.0 u:u.oon~n~¢ o—uuou hum-a no» access: @— Ouanh 83 Table 17 Culling Summary. Eight Pilot Herds in FAHRMX Retrospective Sample Mean Mean Mean Mean Age At Total Mature Days In Reason for Calving Milk Equivalent Milk 011A Cull Code calling 7 Culls (Months) (Pounds) (Pounds) At Culling 30 Sold for dairy 3O 41 ‘ 8816 16119 192 purposes (25)sd (5117) (2539) (122) 31 Sold because of low 67 52 11000 15534 222 production (22) (5434) (3224) (94) 32 Sold because of 12 75 10135 18888 181 physical injury (29) (6232) (4482) (114) 33 Sold because of 13 47 9353 17917 146 nastitis (18) (6464) (4352) (95) 35 Sold because of ten- 1 24 2218 10064 68 peranent 37 Sold because of 18 98 14617 16730 241 sterility (30) (5310) (3147) (90) 38 sold because of old 7 134 12245 15822 228 age (24) (4254) (1996) (79) 39 Sold because of hard- 2 47 7859 16327 147 ware disease (28) (756) (4528) (69) 40 Died because of eilk l 88 19734 19344 337 fever 47 Died because of pneu- 2 101 10280 14201 227 Ionia (52) (4159) (2476) (161) 49 Died because of calving 1 111 --- 17882 352 trouble 50 Sold for unknown reason 8 53 13817 15943 255 (4476) (3158) (66) 52 Sold because of ill- 3 81 15353 18408 192 ness (45) (3277) (6522) (151) 53 Sold because of udder 5 41 5224 15674 89 probleas (16) (2887) (1592) (33) 54 Sold because of leg 6 94 13697 19083 189 probleas (28) (2743) (1158) (37) 55 Sold because slow 1 22 1907 14264 45 uilkcr 58 Sold because 0! dis- 4 69 10452 15838 188 placed aboaasun (40) (8453) (3025) (149) 61 Died because of un- 4 100 15046 18191 336 known cause (49) (7613) (3876) (158) 62 Died because of uastitis 1 92 13659 19144 171 63 Died because of displaced 1 49 665 23057 15 abonasua .sd - one standard deviation. 84 Table 18 Single Equation Run Without Culled Cows Dependent Variable Degrees of Freedom F-Ratio = 28.54 R-Square = .6267 Pounds of Milk per Lactation 272 Significance Level of F-Ratio = .0001 Parameter Standard Significance Variable Estimate Error Levelc Intercept a -479 2129 .8221 Age of Calving b 61 7.88 .0001 Spring Calving (0,1) -953 482 .0493 Summer Calving (0,1) -1234 S39 .0228 Fall Calving (0,1) 722 S67 .2033 Herd 2 (0,1) -400 733 .5850 Herd 3 (0,1) -576 564 .3080 Herd 4 (0,1) 4960 1265 .0001 Herd 5 (0,1) 889 616 .1500 Herd 6 (0,1) -1876 827 .0242 Herd 7 (0,1) -1614 1193 .1771 Herd 8 (0,1) 638 687 .3541 Total Days in Milk 2 SO 31 .1124 (Total Days in Milk)3 .0025 0.133 .9849 (Total Days in Milk) -0.000048 0.00016 .7686 Metritis (0,1) -221 546 .6866 Cystic Ovaries (0,1) 1073 743 .1494 aAge in months. b(0,1) indicates a zero-one categorical variable. The model must explicitly include one less than the total number of categorical variables in each category. cThe larger the number is in this column, the lower the signficance level. 85 average change in each variable holding all others constant. With a low significance level, the parameter estimates can be accepted with a high degree of confidence. Assuming that OLS assumptions hold, the parameter estimates for age of calving and herd 4 can be accepted with 99.99 percent confidence. Age at calving has been shown to have an important influence on milk production (Miller et al., 1970). The herd variables account for variation in milk production between herds. Given the vast differences in the quality of management and animals which exist between herds, it is not surprising to see such differences in herd parameter estimates. Season of calving has also been shown to be a significant source of variation among milk production records (Miller et al., 1970). Miller et al. (1970) found that summer calving was associated with lower milk production, especially among older cows. The negative parameter estimate for summer calving in Table l8 supports this conclusion. If the model is specified correctly, it can be accepted with 98 percent confidence. Hansen et al. (1979) found that health costs were highest during the summer and that these costs were primarily associated with mammary and respiratory disorders. This provides some evidence that the lower milk production records associated with summer calvings may be caused by disease. Of the two disease variables, only the parameter estimate for cystic ovaries is relatively significant. As expected, it is a high positive number, which apparently contradicts the notion that cystic ovaries is detrimental. As explained in Chapter 3, the data set excluding culled cows probably lacks explanatory power. Perhaps the inclusion of culled cows will lower the cystic ovary parameter estimate and more accurately represent the detrimental effects of the disease. Table 19 shows the results of the regression including culled cows. The parameter estimate for cystic ovaries has decreased from l,073 to 940. Because the significance levels are both relatively high (about 85 percent), this reduction of 86 Table 19 Single Equation Run With Culled Cows Included Dependent Variable Pounds of Milk per Lactation Degrees of Freedom 418 F—Ratio = 63.12 Significance Level of F-Ratio = .0001 R-Square = .7073 Parameter Standard Significance Variable Estimate Error Levelc Intercept a 1114 1296 .3906 Age of Calving b 45 5.43 .0001 Spring Calving (0,1) -358 401 .3711 Summer Calving (0,1) -1511 433 .0005 Fall Calving (0,1) 535 487 .2726 Herd 2 (0,1) -765 611 .2112 Herd 3 (0,1) -741 508 .1455 Herd 4 (0,1) 1524 882 .0846 Herd 5 (0,1) -538 532 .3120 Herd 6 (0,1) -2704 675 .0001 Herd 7 (0,1) -1323 925 .1535 Herd 8 (0,1) -167 627 .7900 Total Days in Milk 18 19 .3258 (Total Days in Milk)3 0.19 0.079 .0161 (Total Days in Milk) -0.00033 0.0001 .0010 Metritis (0,1) 127 471 .7872 Cystic Ovaries (0,1) 940 657 .1537 aAge in months. b(0,1) indicates a zero-one categorical variable. The model just explicitly include one less than the total number of cateogrical variables in each category. cThe larger the number is in this column, the lower the significance level. 87 133 pounds is probably meaningful. However, because the estimate is still highly positive, the model does not accurately estimate the loss of production potential caused by disease. Further adjustments, as suggested in Chapter 3, are necessary. Another interesting difference between the two analyses is the change in the significance of the quadratic days in milk parameters. In Table 18 they are very insignificant, but in Table 19 they are very significant. This suggests that non- linear effects of the length of lactation may be more important in culled cows. The drastic change in the parameter estimate for metritis from 127 to -221 should be ample warning not to depend on parameter estimates with low signifi- cance levels. F. Regression Results Including Indices of Genetic Potential Table 20 presents the results from a correlation analysis using the cows in the sample for which genetic indices were available retrospectively. As indicated by the degreeiof freedom (36), the sub-sample is quite small. However, the parameter estimate for cow index is particularly significant, which means it contributes important information to the model. This emphasizes the importance of including genetic information among current FAHRMX data. C. Summary and Conclusions This protocol for the cost-benefit analysis of dairy cattle health management has discussed the opportunities presented by the data storage and analysis capabilities of microcomputers such as those utilized in the FAHRMX project. Microcomputers can be used to reorganize the health management information fed into them. in this application, they are an electronic library-a essentially limited to serving as a herd health reporting system. However, this library of health- related data can also be used for comparative medical purposes--for the cost- benefit analysis of different disease control procedures. Cost-benefit analysis centers around the partial budget, which is in this case an itemization of disease 88 Table 20 Single Equation Run Including Indices of Genetic Potential Dependent Variable = Pounds of Milk per Lactation Degrees of Freedom = 36 F—Ratio = 6.19 Significance Level of F-Ratio = .0001 R-Square = .7451 Parameter Standard Significance Variable Estimate Error Levelc Intercept a 4226 13655 .7565 Age of Calving b 51 22 .0241 Spring Calving (0,1) -2015 ' 1390 .1558 Summer Calving (0,1) -4682 1467 .0029 Fall Calving (0,1) -1231 1270 .3386 Herd 2 (0,1) -3.70 2059 .9986 Herd 3 (0,1) 3235 1562 .0456 Herd 4 (0,1) 4558 2272 .0523 Herd 5 (0,1) 188 2161 .9312 Herd 6 (0,1) 4545 1699 .0112 Total Days in Milk 2 -141 158 .3800 (Total Days in Milk)3 .919 .577 .1201 (Total Days in Milk) —.00127 -l.96 .0575 Metritis (0,1) -1046 1185 .3831 Cystic Ovaries (0,1) 1777 1319 .1863 Cow Index 3.82 1.78 .0387 Dam Index 1.33 1.59 .4092 Sire PD .978 1.22 .4288 aAge in months. b(0,1) indicates a zero-one categorical variable. The model must explicitly include one less than the total number of categorical variables in each category. cThe larger the number is in this column, the lower the significance level. 89 control expenditures and changes in disease impact. Relevant categories have been identified easily enough. However, substantial problems remain in estimating some of these categories-~especially lost production potential. Production potential is lost due to disease in single lactations, across several lactations, and by death and "forced" culling. Therefore, complicated interrelationships exist between disease, culling behavior, and milk production. Using data available previous to FAHRMX, this study was able to utilize disease information on culled cows and thus provide a more realistic data set. However, modeling capabilities were limited by retrospective data. The apparent joint determination of milk production and disease requires the identification of a set of simultaneous equations. Successful identification of these equations depends on the discovery of exogenous variables correlated with each disease but not with milk production. Estimation of the cost and benefit parameters for specific disease control procedures on specific farms will help determine the pay-off from different disease control methods. This is, in itself, a worthwhile objective. It would tell the farmer the optimum return from investment in animal health care, as well as from which control procedure this maximum return could come. Without such detailed information on other farm enterprises, however, the economic value of cost- benefit analysis of disease control is limited. Farmers need to know where their money can best be spent. Investment in animal health care should not preclude investment in a more profitable farm enterprise because of lack of information. This is an argument for whole-farm modeling, parts of which exist today in various forms. The emphasis of FAHRMX on animal health care is due to the presumed high returns from investment in it. The scope of animal health care is very broad, being affected by many aspects of dairy farm management. Therefore, the successful modeling of disease control would be a substantial step towards modeling the whole dairy farm. 10. ll. 12. 13. 14. 15. 90 LIST OF REFERENCES AABP Newsletter. (1979). "Protecting Against Respiratory Disease in Calves." Oregon State University. Reproduced by Guy Reynolds. American Breeders Service (ABS). (1975). Artificial Insemination Manual. pp. 116-48. Anderson, C. R. (1982). Research using daily milk records as a diagnostic tool. Personal communication. Ax, R. L. (1981). "Get 'Em Bred by 15 Months." Dairy Herd Management. Planner issue. Barfoot, L. W.; J. F. Cote, J. B. Stone; and P. A. Wright. (1971). "An Economic Appraisal of a Preventative Medicine Program for Dairy Herd Health Management." Canadian Veterinary Journal 12(1):2-IO. Bhattacharyya, G. K. and R. A. Johnson. (1977). Statistical Concepts and Methods. New York: John Wiley and Sons, Inc., pp. 273-274. Blosser, T. H. (1979). "Economic Losses from and the National Research Program on Mastitis in the United States." J. Dairy Science 62:119-127. Britt, J. H. and L. C. Ulberg. (1970). "Changes in Reproductive Performance in Dairy Herds Using the Herd Reproductive Status System." J. Dairy Science 53:752. Cannon, R. M.; R. 5. Morris; N. B. Williamson; C. M. Cannon; and D. C. Blood. (1978). "A Health Program for Commercial Dairy Herds." Australian Veterinary J. 54:216-230. Carpenter, T. E. and R. Howitt. (1979). "A Linear Programming Model Used in Animal Disease Control." Proceedings from the 11 International Sympo- sium on Veterinary Epidemiology and Economics. Canberra, Australia. pp. 483-489. Coleman, D. A.; W. V. Thayne; and R. A. Bailey. (1982). "High Incidence of Reproductive Disorders in Dairy Cattle." American Dairy Science Associa- tion Annual Meeting and Divisional Abstract. Supplement 1, p. 168. Crandall, B. H. (1975). "Using DHI-EDP Information in Herd Management." J. Dairy Science 58:230. Erb, H. N.; S. W. Martin; N. Ison; and S. Swaminathan. (I981). "Interrelation- ships Between Production and Reproductive Diseases in Holstein Cows: Conditional Relationships Between Production and Disease." J. Dairy Science 64:272-281. Erb, R. E.; S. Wolfe-Selz; and C. E. Coppock. (1975). "Computer Summaries of Life Cycle Data for Cow Research Herds." J. Dairy Science 58:127. Forster, T. L.; U. S. Ashworth; and L. O. Luedecke. (1967). "Relationship Between California Mastitis Test Reaction and Production and Composition of Milk from Opposite Quarters." J. Dairy Science 50:675. l6. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 91 Gittinger, J. P. (1981). Economic Analysis of Agricultural Projects. Interna- tional Bank for Reconstruction and Development. Gould, C. M. (1975). ThelVeterinary Practitioner in Animal Disease Monitoring. D. G. Ingram, W. R. Mitchell, and S. W. Martin, eds. Spring- field, Illinois: Charles C. Thomas, pp. 46—59. Grunsell, C. S.; S. R. Wragg; and J. Allcock. (1969). "The Practicability and Economics of Veterinary Preventive Medicine." The Veterinary Record. January 11, pp. 26-41. Gujarati, D. (1978). Basic Econometrics. New York: McGraw-Hill Book Company. Haller, C. J. (1957). "How to Cut Losses from Breeding Troubles." Successful Farming 55:153-154. Hansen, L. B.; C. W. Young; K. P. Miller; and R. W. Touchberry. (1979). "Health Care Requirements of Dairy Cattle: Response to Milk Yield Selection. 11. Nongenetic Effects." J. Dairy Science 62:1922-1940. Harsh, S. B.; L. J. Connor; and G. D. Schwab. (1981). Managing the Farm Business. Englewood Cliffs, N .J.: Prentice Hall. Hershler, R. C.; C. Miracle; B. Crowl; T. Dunlap; and J. W. Judy. (1964). "The Economic Impact of a Fertility Control and Herd Management Program on a Dairy Farm." J. American VeterinarLMedical Association l45(7):672-676. Hlubik, J. G. (1979). An Economic Evaluation and Replacement Model for. the Lactating Dairy Cow Including Biological Components. Unpublished Master's thesis, Michigan State University. Janzen, J. J. (1970). "Economic Losses Resulting from Mastitis: A Review." J. Dairy Science 53:1151. Kelly, J. W. and J. R. Holman. (1975). "A Modified Herd Reproductive Status Program for South Carolina Dairy Herds." J. Dairy Science 58:261. Kirk, J. H. (1981). "Application of Programmable Calculators to Mastitis Control Programs." J. Dairy Science 64:2048-2058. Kirk, J. H. (1982). Research relating California Mastitis Test scores to DHIA Somatic Cell Count. Personal communication. Lineweaver, J. A. and G. W. Spessard. (1975). "Development and Use of a Computerized Reproductive Management Program in Dairy Herds." J: Dairy Science 58:256. Lipsey, R. G. and P. O. Steiner. (1978). Economics. New York: Harper and Row, pp. 160-161. —_ Louca, Avraam and J. E. Legates. (1968). "Production Losses in Dairy Cattle Due to Days Open." J. Dairy Science 51(4):573-583. 32. 33. 34. 35. 36. 37. 38. 39. #0. 41. 42. 43. 4“. “5. 46. 92 Mather, E.; M. McPherson; S. Harsh; F. Martin; S. Nott; and J. Kaneene. (1982). "Food Animal Health Resource Management System (FAHRMX)-- An Overview." Proceedings from the 111 International Symposium on Veterinary Epidemiology and Economics. Arlington, Virginia. McCauley, B. H. (1974). "The Contribution of Veterinary Service to the Dairy Enterprise Income of Minnesota Farmers: Production Function Analysis." J. American Veterinary Medical Association 165(12):1094-1098. McDaniel, B. ‘I'.; R. H. Miller; and E. L. Corley. (1965). "DHIA Factors for Projecting Incomplete Records to 305 Days." Dairy Herd Improvement Letter 41(6):I-21. Meek, A. H.; W. R. Mitchell; R. A. Curtis; and J. F. Cote. (1975). "A Proposed Information Management and Disease Monitoring System for Dairy Herds." Canadian Veterinary J. 16:329. Meyer, K. F. (1953). "Animal Diseases and Human Welfare." Advanced VeterinalScience 1:1-48. Miller, P. D.; W. E. Lentz; and C. R. Henderson. (1970). "Joint Influence of Month and Age of Calving on Milk Yield of Holstein Cows in the Northeastern United States." J. Dairy Science 53(3):354-357. Morris, R. S. and D. C. Blood. (1969). "The Economic Basis of Planned Veterinary Services to Individual Farms." Australian Veterinary J. 45:337- 341. Natzke, R. P.; R. L. Schultz; G. R. Barr; and W. B. Holtmann. (1965). "Variation in Mastitis Screening Tests and Milk Composition of Udder Quarters Under Normal Conditions and Following Omission of Milking." L Dairy Science 48:1295. Nott, S. B. and G. L. Segwright. (1981). "Future Research on Animal Temperatures Using Electronic Equipment." Los Alamos Scientific Labora- tory, LA-8883, 18 pages. Oltenacu, P. A.; T. R. Rounsaville; R. A. Milligan; and R. L. Hintz. (1980). "Relationship Between Days Open and Cumulative Milk Yield at Various Intervals from Paturition for High and Low Producing Cows." J. Dairy Science 63:1317-1327. , Philpot, W. N. (1967). "Influence of Subclinical Mastitis on Milk Production and Milk Composition." J. Dairy Science 50:978. Poterfield, R. A. and L. E. Heider. (1980). "Regular Programs Lowered Per Animal Health Costs." Hoard‘s Dairyman, July 25. Shanks, R. D.; A. E. Freeman; and F. N. Dickinson. (1981). "Postpartum Distribution of Costs and Disorders of Health." J. Dairy Science 64:683- 688. Statistical Abstract of the United States. (1981). Consumer Price Indexes. p. l167. Thelan, A1. (1982). Director, Michigan Dairy Herd Improvement Association (DHIA). Personal communication. 93 .xou so: 9626 me 533280 05m 65 canon 358 39328 pamoconfimoo .So .5595 me 52:5. 8.: 86.3.3 vein a mo 2:: 05 Sue: $603333 ma amazon-a .565 can: 333688 632523..“ one .35 33qu .352. wheeze-.8 “53:3 mnofimw mo homes a mu one—F >§ozzumk >52 "—0 44¢0 202# ah? 00000 #4 000#¢0 02¢ 00000 #0 000#¢0 02¢ 000.0 #4 .00#¢0 02¢ 00000 #0 .00#¢0 00.01 u000+000.0-000#¢0+.00#¢0-0000#a#2~>4¢0 202# I00 00000 #4 000#¢0 02¢ 000.0 #0 000#¢0 02¢ 00000 #4 .00#¢0 02¢ 0000# #0 .00#¢0 unavv “00000-000#¢0+.00#¢0-000.0a#2~>4¢0 201# ant 00000 #4 000#¢0 02¢ 00000 #0 000#¢0 02¢ 00000 #4 .00#¢0 02¢ 000.0 #0 .00#¢0 unl0v "000.0-000#¢0+.00#¢0-00000a#2~>4¢0 202# a.‘ 00000 #4 000#¢0 02¢ 000.0 #0 000#¢0 02¢ 000.0 #4 .00#¢0 02¢ 00000 #0 .00#¢0 u~a00 "00000-000#¢0+.00#¢0-0000#a#2~>4¢0 202# a00 000.0 #4 000#¢0 02¢ 00000 #0 000#¢0 02¢ 00000 #4 .00#¢0 02¢ 0000# #0 .00#¢0 uua00 “.00#¢0-000#¢0I#2~>4¢0 202# #00 #4 .00#¢0-000#¢0 unah0 u000+00000-00#¢0+.00#¢0-00000a!~0#0# 202# I00 00000 #4 00#¢0 02¢ 00000 #0 00#¢0 02¢ 000.0 #4 .00#¢0 02¢ 00000 #0 .00#¢0 u~a00 n000+000.0-00#¢0+.00#¢0-0000#a8~0#0# 201# av0 00000 #4 00#¢0 02¢ 000.0 #0 00#¢0 02¢ 00000 #4 .00#¢0 02¢ 0000# #0 .00#¢0 u—a00 "00000-00#¢0+.00#¢0-000.0a!~0#0# 202# .00 00000 #4 0w#¢0 02¢ 00000 #0 00#¢0 02¢ 00000 #4 .00#¢0 02¢ 000.0 #0 .00#¢0 u~s.0 ”000.0-00#¢0+.00#¢0-00000a2~0#0# 202# .00 00000 #4 00#¢0 02¢ 000.0 #0 00#¢0 02¢ 000.0 #4 .00#¢0 02¢ 00000 #0 .00#¢0 u—a00 "00000-00#¢0+.00#¢0-0000#a!~0#0# 202# I00 000.0 #4 00#¢0 02¢ 00000 #0 00#¢0 02¢ 00000 #4 .00#¢0 02¢ 0000# #0 .00#¢0 u~a#0 “.00#¢0-0w#¢0a2~0#0# 202# #00 #4 .00#¢0-00#¢0 u~a00 ”000+00000-#0#¢0+.00#¢0-00000a#2~0 201# I00 00000 #4 #w#¢0 02¢ 00000 #0 #w#¢0 02¢ 000.0 #4 .00#¢0 02¢ 00000 #0 .00#¢0 u~a¢0 u0004000.0-#w#¢0+.00#¢0-0000#a#2~0 202# .00 00000 #4 #w#¢0 02¢ 000.0 #0 #w#¢0 02¢ 00000 #4 .00#¢0 02¢ 0000# #0 .00#¢0 u~a00 ”00000-#w#¢0+.00#¢0-000.0a#!—0 201# a.0 00000 #4 #0#¢0 02¢ 00000 #0 #0#¢0 02¢ 00000 #4 .00#¢0 02¢ 000.0 #0 .00#¢0 m~a00 n000.0-#w#¢0+.00#¢0-00000a#2~0 202# a0. 00000 #4 #w#¢0 02¢ 000.0 #0 #w#¢0 02¢ 000.0 #4 .00#¢0 02¢ 00000 #0 .00#¢0 uua0. u00000-#w#¢0+.00#¢0-000m#a#2~0 202# a#. ooo.o .3 pupae oz. ooooo he pupae oza OOOOu .4 .ouhao oza coon» so .ouhao aa-o. ".ousao-»u»«o.»s~o zur» son .3 .umpao-»m»¢o a~.n. uA.a».no.ozc>ozcwpaos:e.ow».uuo.«oz.>oz.u...oo..os.>ozcu»aosae-.agree-n. uAA...ho.hzv>osvupaooae..mpao... “00-0# 004400 0#-'# Ouwuum 0#-00 xw02~0 e0. 00-00 Km02~0 00-00 000m v0-.0 00¢ 00-00 00005 01-00 00> .0-00 000 00:00 002 .0 00-00 0> '0-00 00 00-.0 02 00-#0 .000¢ 00-00 2#0¢4 00-00 .0» 00-.0 .00 a0 o«-m. .oz 5.-.. p» n.... .o n.-u. es o.-o monomno .-v zsoepzoo a.. new: baaz..p "xsexau «baa-o . no z~m>m.m0 202# 0.00000~0 "0.0~#n¢#02 202# . 02 00000.0 u..0u#~0#08 201# .I00000~0 u0.0~#~#0¢I 202# 0 02 00000.0 u..0~#~#0¢2 202# 0.0000000 ”0.4u#0 201# 0 02 00000—0 u..4¢#0 202# 0.0000000 u~.00. Lu.00. uua00. u...0. u~.00. m—.0.. u..0.. u#2~>4¢0\u=4¢>.0—#¢¢4¢>.#.. u.00.\00¢auvenoucauua4¢>.0.. "01.0.+00..e.0.0-u0#0av.00—¢a.0.. ”00...00¢ua\u0uv.u0#0a.w.. n0..8~0#0#.002~0#0#.0.. "0..8~0#0#.002~0#0#.0.. u.44¢u...0200¢00 202# .000 #4 ”.02uua0...0200¢00 202# .0#. #4 02¢ 00000 #4 .00#¢0 02¢ 000.0 #0 .00#¢0 u.¢0#2~3.a.0200¢00 201# .000 #0 00000 #4 .00#¢0 02¢ 000.0 #0 .00#¢0 «en—J‘belcg‘flm ZUIP Aonfl h.— n.802830...0200¢00 201# .000 #4 02¢ 02¢ 000.0 #4 .00#¢0 02¢ 00000 #0 .00#¢0 02¢ 000.0 #4 .uw#¢0 02¢ 00000 #0 .00#¢0 ”.ozaeam...ozomaum 2mm» Ans. .4 02¢ 000.0 #4 .00#¢0 02¢ 00000 #0 .00#¢0 n.¢0#2~3...0200¢00 202# .000 #0 000.0 #4 .00#¢0 02¢ 00000 #0 .00#¢0 n.44¢u...0200¢00 202# .000 #4 ”.002230...0200¢00 202# .000 #4 00000 #4 .00#¢0 02¢ 0000# #0 .Ou#¢0 neg—lfiflslvg‘wm 2!... A“... hJ 02¢ 02¢ 00000 #4 .00#¢0 02¢ 0000# #0 .00#¢0 02¢ 02¢ 00000 #4 .00#¢0 02¢ 0000# #0 .00#¢0 u.¢u#2~3...0200¢00 202# .000 #0 02¢ 00000 #4 .00#¢0 02¢ 0000# #0 .00#¢0 #0 00000-#0#¢0. 02¢ #0 00000-#U#¢0. 02¢ he ocean-pupae. oz. #4 00000-#u#¢0. 02¢ #0 000.0-#u#¢0. 02¢ be ooo.o-.u»¢o. oz. #0 000.0-#0#¢0v 02¢ #0 000.0-#u#¢0 00 0# #4 000.0-#u#¢0.02¢ “.44¢u..#200¢00 202# .000 00000 #4 #0#¢0 02¢ 00000 #0 #u#¢0 u.008200..#200¢00 202# .000 00000 #4 #w#¢0 02¢ 00000 #0 #u#¢0 usguflflmelg‘wm ZUIP A"... 00000 #4 #0#¢0 02¢ 00000 #0 #u#¢0 u.¢u#2~)..#200¢00 202# .000 00000 #4 #u#¢0 02¢ 00000 #0 #w#¢0 ”.44¢u..#200¢00 202# .000 00000 #4 #u#¢0 02¢ 000.0 #0 #u#¢0 ueflwgslg‘um lug aflON 00000 #4 #u#¢0 02¢ 000.0 #0 #m#¢0 ”.02.850..#200¢00 202# .0#. 00000 #4 #u#¢0 02¢ 000.0 #0 #u#¢0 u.¢u#2~3..#200..> .zux.¢mu. no >¢o..o .zua. >asao.44«u .uzom«um no. ¢<> >xsao.uw:.:m .ozom.um no. c¢> >xzao.oz.¢am .uzom¢um no. ¢¢> >xxao.au.z.: 2o..¢.o<4 omocoouc go .umzo no zom¢um..onm¢um mu.u¢>c u..m>u can ¢¢> >xxaouu..m>o. m...u.ux no. c<> >axan.m...u.us n....m<: can c<> >xxaoom....m >xxao.4c.o m>¢u.z. oz.>4¢u..z.>4¢o 4«>cu.z. oz.>4o zo..u:oo¢. 4.2...0..«¢ u.¢o uno >co.ou.«o zo....o<4 .xuz no .umzo no upco.«ou.¢o zo...»a«4 auocoumu go .umzo no u.¢o..uu.«o non 4.2....00. .«u an..:a 44¢0 .00n00a0.00. ”#0—4 201# 0 #4 I~0#0# “#0u4 20:# 0 #4 #8~0 “0.4410 201# .4400. 02 .0200000 "..4415 201# .4440...0200<00 ”0.001100 201# .008300. 02 .0200000 u..008100 20Z# .00!!00...0200000 “0.02.0a0 201# .02uuaw. 02 .0200000 u..02~¢am 201# .02.!00...0200(00 “0.00#2—3 20I# .n0#2~). 02 .0200000 u..00#2~3 20I# .00#2~D...0200(00 ”0.4400 201# 00.004400 00 . .004430 u—.#0. 00.00. L~I00. L~.v0. u~.00. u—I00. 0...0. u~.00. u~.00. u—I0Q. u~.#v. n..4400.0v. ”0.0.0001 201# 0. 08 0001 u..0.0001 201# 0..000I "0.0.0001 201# 0. 02 0002 u..0.0001 20I# 0..0002 “0...0001 Z01# .. 02 0001 u....0801 20I# ...0¢01 ”0.#0¢01 202# # 02 0001 u..#0001 201# #.0002 "0.00001 20.: 0 02 000... u..00¢01 201# 0.0001 "0.0.0001 201# 0. 02 0001 u..0.0¢01 201# 0..0001 «0.00001 ZOI# 0 02 0801 u..00¢01 201# 0.0802 «0.00001 20..# 0 02 0001 ...00001 801# 0.0001 "00.0001 201# . 02 0001 u...0001 201# ..0¢01 3.0.080 20.: 0.08020 8 .0800; u~.00. u~.". u~.0v. u~.0v. m...v. u—IOQ. L~.00. u—.00. 0~.#0. u~.00. u~.00. u~.v0. 0..00. L—I00. u...0. u~.00. L~.00. u~.00. s~.#0. ”0.0.00800.00. n0.u.#0>0 201# 0 02 00000.0 50.00. 123 50 001. 100 0.0. 100. 510. 05 .00- 0.0 .0- 000 10. .00 05000 005 00.: 0.0 50.. .00 0000. 1000. 00.0. 0005. 0100. 10..0 .000. .000. 0000. 0.00. 0000. 000.0 0105. 51.0. 5010. 0000. 5005. .001. 51.0. 0000. 5.05. 0000. 0510. 000.0 0050. 0000 0050 1000. 010.0 0100. 0505. 0.00. 0..00 .100. .000. 0000. 5000. 0000. 00000 0005. 0110. 0.00. 1050. 5005. 1001. 5000. 1..m. 000.. .015. 0005. 00000 10000 .00.00 .000.0 .0.0.0 00..0. .00000 .00.00 001... .000.0 .000.0 .00..0 005... .00000 .0...0 00500. .0..00 .00050 .0.000 .000.0 .00.00 0000.. .00.50 .00.50 005... 00000. .00010 000... 0000.. 000000 .050.0 .00000 .00..0 000000 000... 00000. 00000. 001000 .050.0 0000.. 000.00 .000.0 .00.00 .000.0 001.0. 00000. 000000 .00000 .00000 001000 1000000 GDN’M'NNNDNGDQN "00000 14.2 2. 0001000 14.2 2. 02.0400 000 200100 .0000 4400 1.10.004400 02.4400 000 01) >2200 41>0052. 02.>410\0041>.0.51041> 20.5000000 00 0041) 014400.0041> 0.0 00 >00 ..0\00..00.01.0.0.500\00.00.00.00 0>10 2. 00.000 >00 00 250204 0510 520251005 51 X4.2 2. 0>10.52.0 002000uu.0 0050.0000 00.0.0000.0 41.520500 0.50200 00 X002. 210.X002.0 41.520500 20.515014 5X02 00 50020 00 010>.00> 20.515014 5K02 00 50020 00 >10.000 20.515014 5X02 00 50020 00 15202.002 000000 000000 00 0110 .0.010 05500. 000000 00 0500 .0.010 000..0 000000 00 5500 .0.110 050.0. 000000 00 1500 .0.010 000.00 000000 00 0000 .0.N10 000010 000000 00 50.0 .0..10 000..0 000000 00 0100 .0.010 050.0. 000000 00 .110 .0.000 O00..0 000000 00 0000 .0.000 000..0 000000 00 00.0 .0.500 000.00 000.00 00 1010 .0.000 05..0. 051.0. 00 00.0 .0.000 000000 000.10 00 .500 .0.100 000..0 000..0 00 0000 .0.000 0550.. 001.00 00 0000 .0.000 000.00 000000 00 0000 .0..00 000010 005000 00 0000 .0.000 000000 000000 00 1110 .0.000 0500.. 0550.. 00 5000 .0.000 005.00 00.010 00 0.00 .0.500 050... 000000 00 5010 .0.000 000000 001000 .0 0110 .0.000 000050 000000 .0 0010 .0.100 05..0. 05000. .0 00.0 .0.000 05000. 051... .0 0500 .0.000 "00010..00 0>10 41505.002.0505 .000 0>10 41505.002.0505 .0.0 .0.0 .4400 .5.0 .1.0001 .0.0 .0.0001 .0.0 .0.0001 .1.0 ...0002 .0.0 ..0.0002 .0.0 .0000: ...0 .0000: .0." .5000: .000 .0000: .000 .00001 .500 .1000: .000 ,.00001 .000 .0000! .100 ..0002 .000 .000 ..00 .000 510 005500 5200000.00500 .00. .00.0000 .00. 24.2 2. 0>10 41505 .2.0505 .50. .00. .00. .10. 0.50200 00 K002. 000.K002.0 .00. .00. ..0. .00. 000 00.00 010>.0> .00. 000 00.00 >10.00 .00. to 00.00 2.200.? .50. 20.51.014 00000000 00 50020 00 010»..0> .00. 124 00 050. 050. 551. 105. 00.. 001 105. 00.. ..0 .0 00 .01. 10 .0. 000- 0.0 000 .50 .50 005 0.5 000 000 010 500 011 105 000 0001. .1.5. 0000. 0500. 0500. 00000 0000. 00.1. 0.00. 001.0 10000 01.0. 5005. 0000. 0000. 0.00. 5005. 5000. 10000 .000. 0511. 5..0. 1050. 0110. 00000 1.000 5000. 00100 0005. .011. 0500. 0000. 0005. 0000. 5110. 1000. 0000. 1051. 0050. 5000. 0..0. 0000. 5..1. 0105. 0510. 0005. 0000. .001. 0100. 0000. .00.0 001.0 0005. .001. .10.0 0101. 11.0. 1000. 0.50. 0001. 0510. 5000 0000 0.00 0500 0.00 0000 .000 0110 0150 0050 1100 0000 1010 0000 0500 0000 0000 0000 5010 0050 5000 1.50 5000 .050 1000 0000 .050 0550 0000 0000 0000 0510 0000 1500 .010 0000 0100 .000 5000 0000 5.10 5050 5150 1000 0100 .050 0000 0050 0000 0010 1000 1000 0050 0.00 0000 0050 0150 .110. 50.5. 0505. 0005. 0500. 0500. 0000. 00.1. 11000 ..000 01.0. 00.0. 0000. 5000. 00000 00.0. 0100. ..000 0511. 5000. 1050. 50000 00.10 1.000 5000. 00100 0.0.0 0055. .050. 0550. 0.0.0 0000. 50000 .050. 0000. .010. 0050. 01.0. 0000. 0000. 5..1. 0105. 0510. 0.55. 00.00 .000. 00000 050.0 00000 00.00 .000. 1001. 00000 0100. 0050. 000.0 5.00. 5.50. 01000 .00000 .00000 .00050 .00000 .00.00 .00.00 00000. .00000 .00..0 .00.00 .00000 .00.10 .00000 .00000 .00000 .00.10 .01.10 .01.00 000000 .00000 .0..00 .0.000 .01010 .00000 .00.00 .00000 .00000 .00.00 .00000 0010.0 000000 .00010 .00000 .0.010 .00.00 .0.000 .00010 .00.00 .00000 .0.050 .01010 .00000 .00..0 .0.010 .00.00 .00.00 .01000 001... .00.00 .0.000 .000.0 .05.10 00000. .01.00 .00000 .00010 .00..0 00000. .00.00 .00000 .00000 .00.10 00000. .00010 0000.. .00000 .05.10 .000.0 .00.00 .00.10 .00000 .000.0 .00000 .00010 05000. 00000. .0.000 .00.00 .00.00 .00.00 .00.00 .00..0 .00.00 .00000 .0.000 .00.00 .01.00 .00000 .00... .00.0. .00.00 .00.00 .00.00 .0.0.0 .00.10 .00.00 .00000 .00.10 .00000 .0.0.0 .0.0.0 00000. .0.000 .00.00 00100. .0..00 000000 .00000 .000.0 000000 000.0. 005... .00000 0000.. .0.0.0 .00.00 *Q1-1’v(‘vw.e'flflb.”9')”(Q‘Hngfiarw1D.-01-v-w(”6"(Qe'cu.1[‘wp')fl¢‘FHPI,'IDFD'WPV'OCVGH'1"‘afl,'lnufa’ 000000 000000 00...0 00.050 000000 000.50 001050 000.00 000.10 000000 000.00 000000 000.00 005.00 000.00 000.10 000.00 000010 000010 00.000 000000 000.00 000010 000010 00..00 001010 00.000 000.10 005000 005..0 00.000 000.10 .00..0 00.00. 000010 000000 000000 00.000 000000 001000 001000 001000 001000 000.10 000000 000..0 000.00 000000 050.0. 00.000 05.00. 000.00 000000 051.0. 000000 0510.. 05000. 0500.. 05500. 000000 00.000 000... 000050 000050 000000 000000 00.000 05.000 8888888888888 88888888888888888888888888 .000 55.0 0000 0000 .000 5.00 .000 05.0 0000 00.0 0000 1000 0500 00.0 00.0 00.0 .000 10.0 00.0 00.0 00.0 0000 0000 0010 0010 1000 0000 0000 0010 0500 0500 0000 0000 0000 0500 5500 0500 0500 0000 0000 .000 5.00 0000 0000 5010 5000 0000 0500 0010 1110 00.0 0110 0000 0010 5110 0110 0.00 0..500 0..000 0..000 0..100 0..000 0..000 0...00 0..000 0..000 0..000 0..500 0..000 0..000 0..100 0..000 0..000 0...00 0..000 0..000 0..000 00.500 00.000 00.000 00.100 00.000 00.000 00..00 00.000 00.050 00.050 00.550 00.050 00.050 00.150 00.050 00.050 00..50 00.050 00.000 00.000 00.500 00.000 00.000 00.100 00.000 00.000 00..00 00.000 00.000 .0.000 .0.500 .0.000 .0.000 .0.100 .0.000 .0.000 .0..00 .0.000 .0.010 .0.010 .0.510 125 N5.N 000- 000 050. 0N0 .N. N00. 505 005 505 005 0.0 005 0.0N 00 00.- 00 00 N0 N0 000 000 NNO- 500 000 N0.u 00.: .0- NON 50- 50 00.: 550 00.- .00 000 000. .0.. 00. 000- 00.. N00 00N 000- 00.- .0.. 00. 50. 000 005 00N 000 00N. 0N0.N 0000. N500. .005. 005N. 0.00. 0000. 0000N 0050. 0000. .000. 00.0. 0000. NN00. 0000. 5000. 0005. 0.05. N005. N000. 000.. 0500. 000.. .NNNN 5000. 000N. NN00. 0500. 0000. .055. 0N00. 00.N. 00.0. .000. 0000. 0.00. 0000. N000. 0000. 0000. N000. 0000. 000.. 00N.. N000. 0000. nnmv. no.0. noun. «.00. mm... cvoo. no..n cvuo. sumo. «moan 0000 N000 N000 N000 5000 .050 N000 0N50 0000 0000 0000 0..0 N000 N000 N000 0N00 .000 0000 0000 N000 0000 N000 0000 0N00 0000 0.00 0000 0000 0000 .000 0000 .000 0.00 0.00 0000 N000 N000 0000 0N00 .500 0000 0000 0.00 nwvo 0000 0000 0000 0000 0N00 0000 0000 0.00 0000N 00.0. 0050. 0000. 005N. 0000. 0000N 0.00. 0000. 00.0. 0000. .050. 0000N 000N N000. 0.00. 5.00. 0005. 0000. N000. 0NO0N 000.. 5000. 000.. 0000. 0000. 000N. NN00. 00.0. 5N.0. 00..N 00.N. 00.0. 0.0N. .000. 0000. .0.0N 0000. 0000. N000. 0000. 00.0. 0000. 000.. 00N.. 050N. 0000. 0005. 0000. N.00. 00.0. 0.00. 0000. .00NN NO0NNO .00.N. .05000 NON000 .0..00 .00.N. .00.0. .00.00 .0..50 .00N00 N000.0 N0.000 N000.0 .00... .00000 .00.0. .00N50 .00.0. .00N00 .00N00 .00N00 .00.00 .00N00 .00000 .00000 .00000 .00.00 .00N.0 .00N50 .00.00 .00N50 .00.00 .0..0. .000N0 .00000 .00000 .0NN00 .000N0 .00NNO .0.000 000NN. .000.0 .00000 .00N00 .05.00 .00N00 .00..0 .0.0.0 .00N.0 .00.00 .00000 N000.0 .000.. .000.. .00N0. .00000 .000.. .05000 .0000. .000.. .00000 .05000 .00.00 .00N00 .00NN. N00..0 NON..0 .00.00 .00.0. .0.000 .00.00 .00.00 .00000 .00.00 .00N00 .00.00 .00050 .00000 .00000 .00.00 .00NNO .00.00 .00000 .0.000 0000.. .00000 .0.000 .00N00 .00.00 .0.N00 0000N. 000NN. .00000 .0N0.0 000NN. .0N..0 .0N0.0 .00000 000N0. 000N.. .00N00 000N.. .00NNO 000NN. .00.00 .00.00 0000.. .00.00 005... 00000. .00.00 .0N0.0 '8"fl'fl’NOH’V'D'NNCNNNNN'8""""ONG’O"'“"’N'n"'l~'fl"fl0fl"" 000.N. 000NN. 000.N. 000.N. 000.00 000.N. 00N.00 0000N. .00N.0 00N.0. 00000. 005.50 000N00 00.000 000.00 00.000 00N.00 000.00 000000 000N00 000.00 000000 000000 000000 0000N0 000000 000000 000N50 000N00 000N00 0000.0 000NNO 000.00 000000 000000 00.NNO 00.0.0 000.00 0000.0 0050N0 000000 0000N0 000000 000N.0 000.00 000N50 0000.0 000000 000..0 000.0. 005.0. 000.00 00.000 000.50 000.00 00.000 000.00 000.00 000000 88888888888 8888888888888888888888888888888 0000 cpno «poo .pno nuoo .ouo ouno .nuo none pane o'no oooo ..«o onuo osoo cane cane cane 0000 cane cane ovoo pcno oouo .coo ocuo o.no n«.o .auo nouo oouo ooao oouo poao coao upuo acne .000 ocuo mono o.«o '..o nuuo ..«o .suo case some ooao cpuo ocuo poop «use ..«o us.o choc couo coco snoo 003000 003000 003000 003000 003000 003000 003N00 003.00 003000 003000 003000 003000 003000 003000 003000 003000 003N00 003.00 003000 003000 003000 003000 003000 003000 003000 003000 003N00 0.3.00 0.3000 0.3000 0.3000 0.3000 0.3000 0.3000 0.3000 0.3000 0.3000 0.3.00 0.3000 0.30N0 0.30N0 0.35N0 0.30N0 0.30N0 0.30N0 0.30N0 0.30N0 0.3.N0 0.3000 0.30.0 0.30.0 0.35.0 0.30.0 0.30.0 0.30.0 0.30.0 0.30.0 0.3..0 000000 000000 00 00.0 0.30.0 0 00.000 000000 00 0000 0.3000 0 000..0 000000 00 0000 0.3000 U6 505 V50 000: 0V0 VVV 0V0 0V0 0V0 VVV VVV VVV 5VOV0 0V0 0V0 0'0 0V0 .0 00V. 0V5 00V. .00 000. V00: 00 00V 00 V.0. 000- 000 00 000 0: 00. 55.- .00 00 005- 005- 005- 00V 000- .0- 00. 0V0 00 00 000 000 V05- 000 0..- 00.- .0.. V0. 000: 05. 000 V00 5V0 000. 050 550 00 .00 00V .0. 000- 00. 000 00.. Kn! 000 000 .00 000V. 5000. 0500. .0VV. 0000. 0.00. 0000. 0000. 000V. 0000. 5.000 0000. 00.0. 5.00. 0000. 0550. 00000 00V0. 0V500 0000. 00000 0500. V5.5. 0000. 505V. 0000. 0005. 00V0. 0050. 05000 0550. 000V. 000V. .V00. 0000. 0000. 00V0. 00... 0000. 0000. 05V0. .000. .000. 0000. 0000. 0000. 0050. 0V000 0000. 0V05. 0000. 05.0. 500.0 5000. 0050. 0000. 0000. 50V00 001V. 0000. 5000 5000 000. .050 .000 0000 0000 V.50 0500 0000 0000 0500 0000 .500 5000 .00. 0000 5050 5050 0000 0000 V050 0000 0000 0050 00V0 V000 5V00 0VVO .000 5000 0000 0.50 0000 0VVO 0V00 55V0 V000 0000 0000 0050 0000 V000 .000 000. 0500 0V50 0050 0000 0500 0000 0100 0050 0000 0000 000V. 5.00. 00500 0055. 00VV. 0.00. 0000. 00000 000V. 0000. 5.000 5000. 0005. 0000. V005. V000. 005V0 .000. 00000 0000. 0...0 0055. 00000 000V0 505V. 050.0 0005. 00'0. 005m. .0050 0550. 000V. 000V. 0000. 0000. V000. 00V0. 00... 0000. 0V00. 05V0. 0V00. 0005. 00000 .000. 0V00. 0.0V. 000V0 0000. 0V05. 0000. 0V00. 0000. 0000. 0000. 00000 00VV. 0..V. .0.... .000.. .00000 .00.50 .05000 .05000 .00.00 .0.0V0 .00.00 .0.000 .00..0 .00000 .05000 .00050 00000. .0VOV0 .000V0 .05.00 .00.V0 .0V050 .00.00 .00050 .0000. 00.00. .00.V0 .00000 00000. 0000.. .0.050 .00000 .00050 .00000 .0V.50 .00050 .00050 .00.50 .05.50 000... .00050 .0V.V0 .05.00 .0000. 000000 .0V00. 000.00 .00050 .0000. 00.000 .00000 .0.000 .00.50 000000 .0000. .00000 .05000 000000 .0.000 .00000 00.00. .00.00 000.0. .00.00 .00.V0 .000.0 00.00. .00000 .00000 .0.0V0 000000 .0V000 .00000 .0V000 00.00. .00.00 .00000 000000 000.0. .00.00 .00000 000.00 .0.000 0050.. 0050.. 000.0. .00000 .05000 .00000 .05000 .00.00 .00000 .00000 .00000 .00.00 .00000 005.00 .00000 .00..0 .00000 .000.. 00.0.0 .000.. 0000.0 .00000 .00.00 .00.0. .00000 .05050 .00000 .0.000 .00.00 000..0 .05.00 .000.. v00v 'N'0"'NOUNDI’"'NNOD'GD'Q'M'QN""Nfll’t"fl'fl'flflfl"fifl"’«N00? 0000.. 000... 000... 005000 000050 000.00 000050 000..0 0000V0 000.00 0000V0 05500. 000000 0000V0 00.000 05.00. 0000.0 000.50 000.V0 00V.00 000000 000.50 00000. 05V0.. 000000 000000 0500.. 000.00 000..0 05000. 05000. 00.050 000000 000.00 000050 00.050 000000 000000 00..00 000000 000000 050000 000000 00V0.0 000050 .000.0 .05000 000.0. 00V00. 000.00 000.0. .050.0 000.00 000000 000.50 000.50 000000 .325 0050.. 000.0. 00000. 00000. 00000. 00.000 000000 000000 .00050 000000 000000 000000 0000.0 000.0. 05V.0. 000050 .000.0 000000 8888.8888888888833653888888888888863688888888888888888888 .0 '3" ¢3C>0 0000 .000 00V0 V000 5000 00.0 00.0 0000 ..00 V000 0000 0.00 00.0 0000 0000 0.00 00.0 V0.0 00.0 .0.0 V0.0 0000 0000 0.00 0000 05.0 00.0 0V.0 V5.0 05.0 05.0 .0.0 0000 V500 50.0 .000 0..0 0000 00.0 0..0 0000 00.0 0V.0 00.0 5000 0000 0000 0000 V.00 0000 0000 00.0 5000 .000 0000 0000 .000 0000 0500 0.300V 0.I00V 0.I50V 00I00V I00V 0.:VOV 0..00V 0.I00V 0...0V 0.:00V 0.I0.V 0..0.V 0.|5.V 0..0.V 0..0.V 0.uV.V 0.u0.V 0.I0.V 0.I..V 0..0.V 0.I00V 0..00V 0.I50V 0.:00V 0.I00V 0.IV0V 0.I00V 0.I00V ..I.0V ..IOOV ..0000 ..I000 ..I500 ..0000 ..0000 ..IV00 ..0000 ..0000 ..I.00 ..0000 ..I000 ..I000 ..I500 ..I000 ..I000 ..IV00 0OI000 00I000 00I.00 00:000 00I050 00'050 00'550 00I050 00I050 00IV50 00:050 00.050 00I.50 003050 003000 H7 000 050 000. 0.0 000 0. 00. 000 00- .00- 0005. .0000 0000. 0000. 0500. 0000. 0000. 0000. 05.5. 0000. 00.0. 0000. 0000. 0000. 0000. 0.00. 0000. 0005. 0000. 0000. 0.00. 00.00 00.00 .050. 000.0 0000. 0000. 0050. .000. .000. 5000. 0.500 .0000 .000. 5.00. 0000. 5005. 5.00. 0000. 0050. 0000. 0000. 00000 0005. 00.0. 0000. 0000. 0000. 0000. 0000. 0.00. 00.0. 0550. 0000. 0505. 0000. 0550. 0005. 5000. .000. 0000 0000 0000 0.00 0000 0000 .000 .500 0000 0500 00.0 00.0 0500 5000 5000 0000 0000 0000 0000 0000 0050 0500 0000 .500 0.00 0000 0000 0000 0000 0000 0050 00.0 0000 0000 5500 0000 0000 0000 0000 0.50 0000 0500 0000 ..00 5000 0500 5000 0500 0050 0000 0050 0000 0050 5000 0000 500.. 5505. 0.00. 0.000 0050. 050.. 000.. 000.0 0000. 0005. 00000 00000 0000. 0000. 0.000 0550. 000.. 0000. 0500. 0000. 0000. 005.0 0000. 00000 5000. 0005. 00.0. 0000. 0000. 0000. 00000 0500. 000.. .000. 00000 0505. 0000. 00000 0000. 0.00. 50050 0000. 00050 50000 00050 0000. 0000. 0000. 0050. 0000. 0050. 0.00. 0005. 5.00. 0000. .000.. .0000. .000.. .00... 00000. 00.00. 000.0. 00000. .00.00 00000. 00000. 000000 00.00. .05.00 00500. 000.00 .00000 .00..0 00000. .00000 .00.00 00.00. 0000.. 000.50 .05.00 .00000 .00000 .000.0 00.050 000000 .00000 .000.. .00.0. .0.00. .00000 .00.00 .00000 .00000 .00.0. .00.0. 0000.0 .00000 .000.. .00.0. .00050 .0.00. 00000. .00000 .0.000 .050.0 .0..00 .00.50 00000. 000000 .00000 .05000 .00050 .05000 .05.00 .05.00 .05.00 .05000 .00000 .05000 .00.00 m 0 0050050000000000 N U .N'Nflflflfl" 9g 000..0 00.000 000000 000000 000000 000.00 000.00 000050 00.0.0 000.00 000050 000000 000000 000000 000000 000.00 000.50 000000 000000 050.0. 000.00 00..50 83.. . 005000! 000.00 000000 000.00 000000 0050050 .05000 .00000 .000.0 .00000 00.000 00..0. 00000. .0.0.0 .00..0 .00000 .0.000 .00000 00.00. .0.000 00000. 000... 000.00 000000 000000 00.0.000000 000050!000.00 000.50 000000 .00000 .0.000 .00000 .00000 .00000 .000.0 0000.. 000.00 000.00 000.00 000.00 000.50 000.000 000... .00.00 .00..0 0000.. .00.00 .00000 .00.00 000... 0000.. 000... .000.0 00.... .0.050 .00050 .00.00 00000. 63338888888888888888888888888888888888888 88888888888 0000 .000 0000 50.0 0500 05.0 05.0 00.0 0500 0000 5000 0000 0000 0000 0000 0000 5000 50.0 00.0 00.0 .0.0 0000 0..0 5000 0500 50.0 50.0 0000 .5.0 00.0 55.0 0000 0500 0000 0000 00.0 5000 00.0 5500 00.0 0500 00.0 00.0 0000 00.0 0000 0000 0.00 0000 0000 0000 0000 0.00 0.00 0.00 0000 0000 .000 0000 0.I000 0.I000 0.0000 0.I500 0.I000 0.I000 0.I000 0..000 0..000 0.I.00 0.I000 0..050 0.I050 0.0550 0.0050 0.I050 00.050 00I050 000050 000.50 00.050 00.000 00|000 00.500 00.000 00I000 003000 00.000 00.000 00u.00 00.000 00.000 00I000 003500 00I000 00I000 000000 000000 00.000 00|.00 00.000 000000 000000 003500 00:000 00.000 00.000 00.000 000000 000.00 0.I000 0.I000 0.I000 0.0500 0.0000 0.0000 0.I000 0.0000 0.-000 0.I.00 0.0000 128 00 50 00 3 5000. 0000. 0050N 0N.0. 0000. 0000N 0000. 0000. 50N5. .005. 500.N 0000. .NONN 0.00. 0.05. 0050. 0100. 0550. 0000. 0010. 1000. 0.00N 0N00. 0.N.N .0N0. N050. 0.N.N 005.. N000. 010N. 0000. 1055. 0105. 0501. 0500. 1000. 0N00. 0100N .005. 00000 N10N. 0.5N. 0000. 0000N 000.. 50.0. N.00. .005. 01NN. 00.0. 1050. 5510. 0000. 0010. 0510. 5005. 10.1. .N00. 000N. N.10. 0000N 0000 0000 5500 00N0 05N0 0000 0.00 0N50 5N50 050N0 N1000 0.5NN 1N050 05000 00.0N .000. 0N10. 5005. 01.0. 0000N 1NO0N 50NON 0.00. 0NON. N000. 00010 N0000 05.0. 00000 10N00 N000. 10000 050NN N0000 1010. 050NN ..000 05000 N00.. .000. 00.0. .100. .0N1. 0050. .500. 1N10. 0000N 0011. 00000 00500 1.500 0501. 5000. 0010. 010.. NON0. 0051. 00000 000.. 000N. N000. 05000 05000 1.01. .NN00 .N5.. 0N01. 00000 N1.0. 000.N 005N0. .00..0 .00050 .00..0 005N0. .00000 .00N00 .00.10 .00.00 .00N.0 .00NNO .00N.0 000.N. 00.... 005N00 000.10 000N00 0000.0 000.10 001NOD.000.00 000.N0 000.10 005... 000.50 000.N. .0N..0 .00000 000.N. 000.00 00N000 .05N.0 .0.000 0000N. 000... .0NON0 000.N. 000.50 .00000 .00NNO 000N00 .05N00 .0NON0 .01NNO .00000 .00000 .00.00 000NN. .00.00 .00.00 .05.00 .0N0.0 000NN. .01NNO 00.0N. .0N0.0 00.N0. 000N00 00.0N. .0N0.0 00500. .0N0.0 .00N00 0N . 000.00 00 N 000.00 50 1 001.50 5N . 005050 01 N 001050 00 0 000000 0N . 005N00 N0 0 000000 00 N 000010 00 0 000.00 N0 0 005.N0 00 0 0050N0 50 1 0000N0 .0 0 050N.. 55 0 050N00 50 0 05..0. 0N . 05000. 0N . 05N... 10.0 051.0. 50 N 05N00. 0N . 05N.0. .0 N 000000 50 1 000010 00 0 00N..0 50 . 000.0. 00 N 001N10 00 0 00N..0 .0 . 000.00 00 . 000010 0N . 001NNO 10 . 00.000 N0 N 00.000 00 N 005NNO N0 0 00.000 00 0 00.0.0 .5 1 0000N0 .0 0 000000 01.0 005N.0 N1.5 0550.. 00 . .00.N0 50 0 .000.0 00.5 000.00 50 0 000N10 No.0 00.N00 0..0 005.00 N0 N 00N000 11 N 001.50 01 N 000050 01 N 005000 01 N 00NN00 00 . 000010 00 . 00..00 .1 N 000N00 10 0 001.50 01 N 000000 .1 N 000010 00 0 000000 01.0.000N00 00 . 000010 01 N 000.00 00.5 000.00 000N.0 005N0. 000.00 000.00 0000N0 .01.N0 8888888888888888888883888888888888888688888888888888888888888 0010 0N10 0NNO .010 0N10 .0N0 0010 0NNO 00N0 NON0 00.0 05N0 51N0 05N0 1NNO 00.0 N010 1010 00N0 05N0 0010 00N0 .0N0 0.00 0000 0100 0.00 0000 0000 N000 0000 0100 0000 1000 0.00 0010 1010 N5.0 N0.0 1.N0 .500 1000 N..0 0000 0000 00.0 00N0 .0N0 01N0 51N0 55N0 00N0 N1N0 00N0 0.N0 5.N0 00.0 00.0 05N0 00N0 0000 .03.00 .03000 .03010 .03010 .03510 .03010 .03010 .03110 .03010 .03N10 .03.10 .03010 .03000 .03000 .03500 .03000 .03000 .03100 .03000 .03N00 .03.00 .03000 .030N0 N030N0 N035N0 N030N0 N030N0 N031N0 N030N0 N03NN0 N03.N0 N030N0 N030.0 N030.0 N035.0 N030.0 N030.0 0.31.0 0.30.0 0.3N.0 ..3..0 ..30.0 ..3000 ..3000 ..3500 ..3000 0.3000 0.3100 0.3000 0.3N00 0.3.00 0.3000 0.3001 0.3001 0.3501 0.3001 0.3001 0.3101 0.3001 0.3N01 0.3.01 129 .0 00 50 00 50 00 N0 010.N 050.N 00.1. 5155. 0100. .0N0. .0150 0000. N510. 0015. 0000. .010. 000N. 5500. 0000. N510. 0000. .N0.N N51.N 1500. .100. 0.00. N050. 00N0. .000N 0500. 5000. 100N. N01.N 0.01. 0000. 1001. .510. 1005. 0015. 1000. N055. 0005. 00NON 0.05. 0000. 50.0. 5550. 1000. 0000. 0100. 0000. .050. 0555. 0500. 0010. 0.00. ..00. 0000. 0.10. N0.0. 0N00. 0001N N000 5000 0000 0000 05 0000 0N.0 0500 00.0 0000 0000 N000 1100 1000 0000 0N10 00.0 0.N0 0000 1N50 5000 0050 0000 1100 N000 0.0. 0N.0 N510 0000 0000 0000 0N00 0010 0000 0050 0N00 0500 N050 1150 N.50 1500 00.0 05.0 0000 N0.0 .000 0000 0000 .000 0000 0000 0000 .000 :3 $30 0000 coco 0.00 osno 0NON. 000N. N10N0 05000 NO0N0 5..N0 010N0 000NN 05N10 000.0 005.0 000N0 0150. 000N. 0N00. 0000. 05N10 00.00 0001. 000.N 1100. 0000. N005. N050. 00000 0..00 00N0. 01000 100N. N01.N 0000. 10N0. 01000 0000. 0505. N000. 01.00 N055. 0..0. 0000N 0000. 0.0N0 .0000 0.010 .100. 0.NNO N00O00 NO0NNO .000.. .0ON50 .0NN00 .00.00 .00000 .00N00 .00.00 .05000 .01000 .0..00 .00.0. .0N050 0N100. 0500. .000. 5.50. .000. 01N0. 0010. 0.00. 0.0.0 0010. 0000. 0050N 150ON 10000 0.00. .00050 .00.0. .01N00 .00N10 .00.00 .00.00 .0N.50 .01.10 0010.. 000N0. .00.N. .01NN. .00.N. .00.N. .01N.. .0..00 .00N50 .00N50 .00N50 .00N50 .01N.. N01..0 NON..0 .00.0. .00N50 .01N00 .0..10 .00.N. .000.. .00... .00N.. .0N... .00N50 .00.00 .00.N. .01N00 .00.00 .00.00 .00000 .0N..0 .0N..0 .0.000 .05N00 .00.00 .05N.0 .05010 .00.00 .00000 .00.00 001N0. 000N0. 000N0. .0.000 .00000 .0.000 .0.N00 .0.000 .0N..0 .00000 .0.000 .00010 .0.000 000N00 .00.00 .000.0 .0.000 .05.N0 .05000 .05N1o 00 0 000N00 .1 N 001N00 0N . .0.0.. 51 0 .0..00 0N . .000.. 0N . .010.. 00 1 .0NN00 01.0..00000 N1 N .00000 01 N 000.50 0N . .05.00 0N . .01000 N0 0 .00N10 .0 . .0NN00 N0 . .00000 00 N .0N010 N1 N .00N00 50 0 .00N00 00.0.000.00 10 1 .01000 1N . .0.N.0 01 N 001NN. 1N . .010.0 50 0 .0NON0 10.5 00.NNO N0 0 0000N0 0N . .01NNO 0N . .00..0 01 0 000000 No.5 000.0. 0N . 000N50 00.0 000N00 00.0.00N.50 0N . 000N50 00 N 001NN. 0N . 000000 50 0 000000 05 0 000000 .1.0.00.00. 5N . 00NNO. N0 0 000050 0N . 00..00 01 N 000000 00 0 000.00 15 1 .00..0 1N . 00.00. N1 N .0N0.0 00 N 000.50 00 N 005N50 00 1 001.50 05 1 000000 0N . 000.00 00 N 000.00 10 1 000N00 10 . 00..10 0N . 005.00 00 1 00..00 50 1 000.00 00 N 000.00 00 1 .0.0.0 00 1 001... 88888888888888888888888888888888888888888888888888888888888 00N0 5.00 ..10 0N00 N.10 0.10 0000 N000 5100 .000 0010 0000 00.0 0000 N000 0100 5100 0.00 0500 0000 0000 0500 0500 OONO 00.0 .0.0 .000 .000 0000 01.0 0000 0NNO 0000 5000 0N00 0100 05.0 00N0 0000 5000 .000 N000 .N00 00N0 0000 0000 0500 0000 0000 5010 5010 0000 5000 0.00 5000 0500 .000 0010 0110 00 55N0 00 05N0 003N.0 003..0 0030.0 003000 003000 003500 003000 003000 003100 003000 003N00 003.00 003000 003000 003000 003500 003000 003000 003100 003000 003N00 003.00 003000 003000 003000 003500 003000 003000 003100 003000 0O3N00 003.00 003000 003050 003050 003550 003050 003050 003150 003050 003N50 003.50 003050 003000 N03000 N03500 N03000 N03000 N03100 003000 N03N00 N03.00 N03000 N03000 N03000 N03500 N03000 N03000 .03100 .03000 .03N00 130 .0 0000. 55.0. 05N5. N0.5. N000. 0..5N 0000. 0000. 055N. 0015. 000.. 0101. 0N0.. 5500. 0010. 0000. N00.. 00050 0111. 5000N .0000 0000. 100N. 0.50. 0000. 0.10. 0001. 055N. 10.0N .00.N ...0. 1100. 0.10. 0000. 005N. 00NN. 1001. 0000. 00N0. 0000. 100.N 0000. 1.00. 0105. N100. .0N0. N100. .500N 1000N 1000. 0000. 0100N 0005. 0100. 01N.. 0010. 0000. 00.0. 001.. 1500. 55N0. 5000 0000 .000 0000 N000 5050 00N0 0050 00N0 01N0 0N00 00N0 .000 «.10 .mno mono 00.0 0000 0000 01N0 0000 NN00 0100 .000 00.0 5000 1000 0000 0010 1010 0010 5110 0000 0000 0.50 0000 N000 5010 1010 5000 0010 0N00 1050 5500 .000 1500 00.0 05.0 N.N0 1010 0.00 55.0. 0005. 00.NN .000. 0.05N 0001. 0000. 50000 1005. 00.00 .0N50 N.000 00000 000.0 0000. N00.. 00050 N0050 00000 1N.N0 0N00. 000.. 15050 0000. 0.10. 0000. .0000 1NON0 0000. 0.10. 010.0 005N. 00N0. 1001. 0000. 0000. 1..0. 100.N 0000. 1.00. .005. 0100. 0N10. N100. 50..N ..00N 1000N 10NNN 05000 0100. 00500 10010 5N.00 00000 N00.. 0000. .000.0 .0000. 00..0. .01.0. .05010 0010.. 000000 .05.00 .0N0.. .00.00 .0..0C .00.10 .01.50 .0.N00 .00N50 .00050 .00N00 .00.00 .0NN00 .00000 .0.000 .0N.00 .05.00 .00N00 .00N00 .0.000 .00.00 .0..50 .000N0 .00N10 .00N00 .01.00 .00.00 .00.N0 .00000 .0N0.0 .00010 .00000 .00.00 0000.. 00.000 000.00 .00N00 .00.N0 00.00. 00NN50 00.00. 000000 .00.00 .00N00 .00N00 .05N10 .00010 a 00000 n.0no .rqnno CenN.. 000.00 00N.50 000050 .00.00 .00.00 .00000 .0NN00 .0NN00 .00.00 .00.00 .00NNO .00000 .05.N0 .00.00 .00.10 .00.00 .00.10 .00.00 .01N10 .0.010 .00.10 .05.10 .0.N10 .01.10 .00N00 .0..10 000NN. .01NNO .01000 .000N0 .00010 .00N10 00.0w. .0.010 005.00 000N0. 000ON. O00NO. 00000. 00500. .05010 .00N10 .00N00 00 . 05100. 00 0 050.0. 5N . 05.0.. N0.0 001N00 00 0 000000 00 1 0500.. 5N . 05.N0. 05 0 001.00 00.0 0550.. 01 N 00.... .1 N .00..0 5N . 000NN. 00 N 000NN. 00 . 000.N. N0 1 000.00 00 1 00.N10 00 1 000.00 01 0 000N10 10.5 000010 01 0 005N00 00 0 000000 00 1 000N00 01 N 00N000 0N . 000000 50 0 00N.00 1N . 000N00 5N . 00N.00 00 N 001000 00 0 .0.N.0 00 N .05N.0 0N . 00N.00 1N . 000.00 1N . 000N50 5N . 000N00 1N . 005.00 0N . 000050 5N . 001N00 00 5 000.00 00 N 00..00 00 1 00.N00 00 N 001.00 00 0 00NN00 00 0 000000 00 N 000N00 00 1 00.000 N0 1 000N10 00 1 00NN00 00 1 000010 00 0 000.00 N0 0 0000N0 01 0 000N10 00 1 00NON0 1N . 000N.0 05 1 000N00 0N . 000.50 .0 . 000.50 01.0.000050 0N . 000000 05 0 00N.00 00 N 000.00 8888888888888888888888888888888888888888888888888888888888888 0.N0 00.0 0.N0 .5.0 00.0 10.0 5.N0 .0.0 00.0 0100 50.0 0.N0 05.0 10.0 0000 .000 0..0 .000 1..0 00.0 .0.0 10N0 0000 0.00 1000 N000 00N0 0..0 N5.0 50N0 NON0 00N0 .000 .0N0 00N0 50N0 50.0 0.N0 01.0 10N0 00.0 N0.0 1NNO 0..0 0000 00.0 1500 .000 .5.0 0N00 05N0 0000 N000 .000 0100 0.N0 0N00 0.3050 0.3N50 0.3.50 0.3050 0.3000 0.3000 0.3500 0.3000 0.3000 N.3100 ..3000 ..3N00 ..3.00 ..3000 ..3000 ..3000 ..3500 ..3000 ..3000 ..3100 ..3000 0.3N00 0.3.00 0.3000 0.3010 0.3010 0.3510 0.3010 0.3010 0.3110 0.3010 0.3N10 0.3.10 0.3010 0.3000 0.3000 0.3500 0.3000 0.3000 0.3100 0.3000 0.3N00 0.3.00 0.3000 0.30N0 0.30N0 0.35N0 0.30N0 0.30N0 0.31N0 0.30N0 0.3NN0 0.3.N0 0.30N0 0030.0 0030.0 0035.0 0030.0 0030.0 0031.0 0030.0 131 .n 00 mm 5' on «@000 0100. v.n¢- 0N5". convo «man, 0000 ooano .O5Nm0 n. « ..0ufl0 —«.0 00090 .Oocv0 5“ . .000.0 0000 v.50. .0..50 .00.00 on . 065050 mono nN5N. .00N00 .00.50 to N Ooouou 0500 500.0 0nn.oo «« . 00.0.0 auno 00050 00'.50 .9.0.000«00 Don... cm a 0.0010 ”wooumwo >0 0'00 "MZCUS Docuunoo ”wooumuo >0 .N00 “530m coca-'00 00 Q.«0 0.3000 00 nnNO 0,-050 00 ONNO 0.3.50 00 ovuo 9.0550 00 oauO «.0050 00 «n.0 0.0050 00 veno n.3v5o 132 APPENDIX 5 CULLING RATE IS DETERMINED PRIMARILY BY THE NUMBER OF REPLACEMENTS THAT CAN BE RAISED To substantiate this assumption, let us work through a hypothetical example. Assume we have a lOO—cow herd. How many calves will we have from which to choose replacements? If we assume that our herd's average calving interval is 13 months, our 100 cows will produce about 92 calves per year if they all calve before being culled (100 cows * l calf I 13 months * 12 months/year = 92.3 calves/year). Of those, 92, 46, or 50 percent will be female. If 5 percent of the females die, about 42 will remain. Of those 42, how many will be successfully bred? If about one-third have breeding problems, we only have 30 heifers that can serve as replacements. Therefore, within this hypothetical herd, we could sustain at most a 30 percent annual culling rate using only our own replacements. Disease and health management problems obviously influence replacement options. For treatment of this subject, refer to the costs associated with extended calving intervals in the text. 133 APPENDIX 6 POSSIBLY PREVENTAPLE CULLS DUE TO DISEASE AND ACCIDENTS, DATA FRCM EIGHT FARHMX PILOT HERDS* Reason for Culling Cull Code Number Culls Physical Injury 32 12 Mastitis 33 13 Sterility 37 18 Hardware Disease 39 2 Milk Fever 4O 1 Pneumonia 47 2 Unknown Reason 50 8 Illness $2 3 Udder Problems 53 S Leg Problems 54 6 Slow Milker 55 1 Displaced Abomasum S8 4 Unknown Cause 61 4 TOTAL 79 Total cull for all reasons = 187. 79/187 = .42 of all culls possibly due to disease and acci- dents. Weighted average culling rate for all reasons a 24% .42 (24%) = 10% of all culls possibly due to disease and accidents. *Derived from Table 17. 134 APPENDIX 7 IS THE PROBABILITY OF BEING CULLED DUE TO DISEASE RANDOM ACROSS PRODUCTION LEVELS? Evidence from Erb et al. (1981) suggests that high production makes cows more susceptible to disease, which would mean that the probability of disease is not random across production levels. However, this same evidence, along with results presented in this thesis, suggests that disease would probably have to be less severe in a low producer to warrant culling. This is because farmers are willing to pay less to maintain a low producing cow. Therefore, higher producers may be likely to get diseased, but low producers are more likely to get culled if they get diseased. Because these two phenomena counteract each other, the assumption of random probability of culling a_ng disease, across production levels, is probably acceptable. The results presented in Figure 14 support this conclusion. The larger histogram shows the distribution of all culls in the eight-herd sample by production level (excluding culls for dairy purposes). Milk production has been mature equivalent adjusted to 305 days in both histograms. The mean milk production of all cows in the sample is about 17,000 pounds of milk per year. The larger histogram is skewed to the left demonstrating the higher culling rate for low producers. The smaller histogram depicts only the culls from Appendix 6, which is a subset of the whole culling sample. The smaller histogram, therefore, represents more mmmmu :0wposw0pm ooccm oocmnfi oocmfi comm~ cocoa comm P 135 Ac xwvcoga< oomv mucocmoom can ommommv on ozv m-zu u asap: uo_ma uzmwm .~o>o; cowuussOHa A: m~qu mo cowusnmhummc «a ocsmaa "unnu— I.o~ 1 cm I av 136 only those animals which may have been culled for disease or accident reasons. Its distribution is less skewed than the larger histogram, which provides some evidence that a higher propensity for high producing cows to get diseased counteracts the increased likelihood of culling low producers because of disease problems.