- Email: [email protected]
Associations between farmers' personal characteristics, management practices and farm performance
Br. vet. ,7. (1990) . 146, 1 5 7
ASSOCIATIONS BETWEEN FARMERS' PERSONAL CHARACTERISTICS, MANAGEMENT PRACTICES AND FARM PERFORMANCE
H . D . TARABLA and K. DODD Faculty of Veterinary Medicine, Department of Farm Animal Clinical Studies, University College Dublin, Ballsbridge, Dublin 4, Ireland
SUMMARY A survey was carried out in a random sample of 123 dairy farms from the east of Ireland . The monthly mean production per cow was 315 1 of milk and 11 . 5 kg of fat . The mean log herd somatic cell count was 5 . 45 (arithmetic mean=372 573 cells/ml), with almost 50% of the monthly counts over 300 000 cells/ml in a 12month period . Bivariate and multivariate analysis was performed to assess the relative impact of the personal characteristics of the farmer and the management policies he applied on the amount and quality of the milk produced . In five out of six models the group of variables related to farmers' attitudes, values, and sociodemographic profile explained a similar or greater amount (between 14 . 44 and 34 . 35%) of the variation of farm performance than the group of management variables (between 14 . 33 and 25 . 99%) as measured by the R 2 . These results stress the importance of the human factors in explaining variation in farm performance .
INTRODUCTION Only part of the variation in farm performance is due to differences in the quality of land, scale of operation, labour and capital inputs . The rest of the variation must be explained mainly by the factor management and the human factors linked to the adoption of management procedures (Muggen, 1969) . Conventional farm management studies have largely ignored the latter, showing no concern for the fact that the enormous growth in agricultural technology must be ultimately incorporated by individual farmers (Frawley, Bohlen & Breathnach, 1974-75a) . They are by no means an homogeneous sector of society (Gasson, 1973), and their personal and social profile affects the choice of techniques to be applied (Jones & Daw, 1964) and farm performance (Frawley, Bohlen & Breathnach, 1974-75a, b ; Bigras-Poulin, Meek & Martin, 1984-85 ; Frawley, 1985) . As far as mastitis is concerned, despite the fact that farmer's attitudes, knowledge and awareness have long had scope for change or improvement (Francis, 1984), studies have largely dealt with the effect of the environmental and management factors on the disease . The objective of this study was to assess the relative impact of the personal characteristics of the farmer and the management practices he applied on the amount and quality of the milk being produced .
158
BRITISH VETERINARY JOURNAL, 146, 2 MATERIALS AND METHODS
A survey was conducted during the summer of 1987 in 123 randomly selected dairy farms from the east of Ireland . Sociodemographic and psychological variables and management variables were collected using a structured questionnaire . In all cases the interviewee was the farm manager and the interviewer the senior author . One week before the interview a presentation letter was posted to the interviewee . Whenever possible, telephone appointments were made afterwards . Each farmer was paid at least five visits on three different days the aim being to reach all farmers selected . Only two farmers refused to be interviewed and had to be replaced . The questionnaire dealt with a wider spectrum than that covered in this paper, and was tested and adjusted three times before the actual survey . Validity and reliability of the answers were dealt with as suggested by Bigras-Poulin et al. (1984-85) . A summary of the variables used in this study is provided in Table I . To measure awareness of subclinical mastitis (AWARE), a simple definition of subclinical mastitis was provided and the farmers were asked to mention ways of detecting this subclinical stage at the cow or herd level . A correct answer was taken as a proxy for mastitis awareness . The research procedures to assess goals and values such as TRADTION, FAMILY, INDEP, PROFIT, and JOB had been developed by Ilbery (1977, 1983), using approaches suggested by Gasson (1973) . HOUSHLD, INFSEEKB, LEVELIV, RISK, and CREDIT were measured using the approaches developed during previous research in Ireland (An Foras Taluntais, 1984 ; Frawley, 1985) . Farm performance as regards mastitis status, milk yield, and quality was measured for the previous 12 months by the following variables : mean log herd somatic cell count (XLOGSCC), proportion of months with SCC > 300 000 (SCC300), mean grade based on total bacterial count (XGRADE), mean milk yield per cow (XCOWYLD), mean milk yield per cow in the month of peak production (PKCOWYLD), and mean fat production per cow (XFATPROD) . Descriptive statistics were generated for all variables . Normality was checked using normal probability plots, and the Kolmogorov-Smirnov D statistic was used to test goodness of fit. To look for associations between farmer's personal characteristics, management practices, and farm performance, analysis of the data was carried out in three stages, and was a modification of the approach used by Bigras-Poulin et al. (198485) . First, bivariate tests for individual associations between the independent variables on each of the dependent variables were performed (analysis of variance and Student t-test for categorical data and correlation coefficients for continuous data) . Only independent variables that were significantly associated (P<0-15) with each dependent variable in this bivariate screening were subsequently used in the multivariate analysis . Scatter diagrams between continuous and dependent variables were also produced . Monthly herd somatic cell counts were transformed to log s and then averaged to obtain the mean log SCC for each herd . Log transformation was used to avoid the problems of nonhomoscedasticity and positive skewness of the somatic cell count distribution . Second, with the independent variable set relevant for each dependent variable, six separate multiple linear stepwise regression analyses were performed . Graphical analysis of residuals and colinearity diagnostics were also carried out . Cook's D statistic was calculated to detect influential observations . Nominal variables were included in the regression equation as k-1 dummy variables to index the k categories of the variable of interest (Kleinbaum & Kupper, 1982) . Third, the R 2 was calculated for a model
cgothv.Aernwmypuasfl
FARMERS' PERSONAL CHARACTERISTICS AND FARM PERFORMANCE
159
(SAS Institute Inc ., 1985a, b) .
Table I Farmer's personal characteristics and management practices under study in 123 dairy farms from the east of Ireland, 1987
Farmer's profile
Management practices
Sociodemographic AGE Age of farm manager EDUCAT Number of years of formal education DEPEND Number of farm dependents HOUSHLD Household structure INFSEEKB Information seeking behaviour
TOTMEET
COUNTY HATOT
HERDSIZE LEVELIV HOLIDAY
EMPLOYEE
Number of farmer's associations meetings attended last 12 months Geographical location Workable land (ha) (owned + rented) Milking Bows + milking heifers + dry cows Level of living Holidays last 12 months Part or full-time
Values and attitudes TRADTION Continuing farming tradition FAMILY Working close to family INDEP Independence PROFIT Maximize income JOB Doing worthwhile job SATIFARM Satisfaction with farming AWARE Subclinical mastitis awareness CONTVIEW Farmer's view of mastitis control LIKEMILK Attitude towards milking INNOV Innovativeness RISK Risk-willingness Use of credit CREDIT
DAIRY
PROPDRY
HEIFMILK
SORCREP
HOUSNG BEDDING SLUROFTN M-IMTYPE MMOLD
TESTOFFN
BULKTANK MILKOFTN STRIP WUDER
WUDERHW HOTWATER DRYCWTH TEATDIP CULL RECORDS IDENTCOW
MILKLAST
WDRWALL
WDRWDAYS
Workable land used in dairying (%) Proportion of dry cows(%) Proportion of first calved heifers in milk (%) Source of replacements
Winter housing type Bedding type Slurry scraping (times/week) Milking machine type Number of years of milking machine Milking machine testing (times/year) Milk cooling system How often the farmer milks (times/week) Foremilk stripping Udder washing (proportion of cows) Udder washing (how) Hot water where the cows are milked Dry cow therapy Teat dip/spray Culling (%) Mastitis records Identification of cows under mastitis treatment Cows with mastitis are separately milked Withdraws milk from all 4 quarters after antibiotic treatment Withdrawal period (days)
1 60
BRITISH VETERINARY JOURNAL, 146, 2 RESULTS
On average, almost half the monthly cell counts were over 300 000 cells/ml, and that figure has to be related to an overall arithmetic herd somatic cell count mean of 372 573 cells/ml (mean log=5 . 45) . SCC300 and XGRADE had different degrees of departure from normality . XGRADE was positively skewed and leptokurtic (had a high peak), meaning that most milk samples were close to the lower values of the scale (0 being premier quality) . Conversely, SCC300 showed very little skewness and was platikurtic (the distribution was relatively flat and with short tails) . The average cow produced approximately 315 1 of milk and 11 . 5 kg of fat per month, while the average cow during the month of herd peak production yielded 3921 of milk. After multiple stepwise regression human variables appeared in all regression models (Table II) .
Table II Multiple stepwise regression : herd milk quality and production functions (variables and P values), Ireland, 1987 XLOGSCC
SCC300 XGRADE XCOWYLD
PKCOWYLD
XFATPROD
DRYCWTH (0 . 001), LIKEMILK (0 . 02), COUNTY (0 . 02), FAMILY (0 . 03),
SORCREP1 (0 -l), WUDER2 (0 . 1) TESTOFTN (0 . 02), LIKEMILK (0 . 03), RISK1 (0 . 01), HERDSIZE (0- l), CULL (0 . 06), BEDDINGI (0 . 07), BEDDING5 (0 . 1) HOUSNG1 (0 . 0001), AWARE (0 . 0001), DRYCWTH (0 . 002), MMTYPEI (0 . 003), WDRWDAYS (0 . 06), LEVELIV (0 . 09), MMOLD (0 . 1), INFSEEKB (0 . 1), HOLIDAY (0 . 1) DRYCWTH (0 . 0001), AWARE (0 . 0009), MMTYPEI (0 . 006), WUDERHWI (0 . 01), TRADTION (0 . 009), HOLIDAY (0 . 02), MMTYPE2 (0 . 05), INFSEEKB (0 . 02), HOTWATER (0 . 06), WDRWALL (0 . 1), TOTMEET (0 . 1) DRYCWTH (0 . 0001), TRADTION (0 . 02), INNOV (0 . 03), WDRWALL (0 . 06), WUDERHWI (0 . 05), HERDSIZE (0 . 09), CULL (0 . 003), LIKEMILK (0 . 09), TOTMEET (0 . 01), HATOT (0- l), MMTYPEl (0- l), INFSEEKB (0 -ij, HOLIDAY (0 . 1) AWARE (0 . 000l), DRYCW'TH (0 . 003), CREDIT (0 . 02), INDEP (0 . 1)
Low log cell counts were associated with geographical location, the use of dry cow therapy, breeding its own replacements, a positive attitude towards milking and working in family . High log cell counts were associated with washing some cows only . Having a high proportion of monthly counts over 300 000 cells was associated with small herds, irregular testing of the milking machine, type of bedding, and low risk-willingness, while low SCC300 counts were linked to a high score in the farmer's attitude towards milking and high culling . Low total bacterial counts were associated with the use of dry cow therapy and high counts with `tide up' housing, old milking machines, bucket plants, shorter withdrawal periods after antibiotic treatment, and low score in information seeking behaviour . High milk yields were linked with bigger herds, the use of dry cow therapy, good information seeking behaviour, high awareness of subclinical mastitis and high culling . Meanwhile, low yields were associated with bucket plants and barn
FARMERS' PERSONAL CHARACTERISTICS AND FARM PERFORMANCE
161
pipelines, lack of hot water where the cows are milked, the use of a single wet cloth to wash the udder, low attendance at farming related meetings, high score in continuing the family farming tradition and no holidays . High fat yields were associated with treating cows at drying off, the use of credit, high independence, and high awareness of subclinical mastitis. Residual analysis did not show any violation of multiple linear regression analysis assumptions. No significant influential observations or collinearity problems were detected . The models explained between 24 . 45 and 66 . 40% of the variation of the dependent variables . The variables addressing farmers' personal characteristics explained as much or more variation of XLOGSCC, SCC300, XCOWYLD, PKCOWYLD, and XFATPROD than management variables (Table III) .
Table III Contribution to the variation on herd milk yield and quality made by human and management variables (R Z ) in 123 herds, Ireland, 1987
Variables Dependent
XLOGSCC SCC300 XGRADE XCOWYLD PKCOWYLD XFATPROD
Model Independent
Personal
Management
14 . 40 15 . 99 25 . 86 34 . 35 30 . 80 24 . 71
14 . 33 11 . 41 42 . 43 18 . 70 25 . 99 11 . 31
24 . 45 27 . 23 66 . 40 50 . 79 47 . 85 39 . 60
DISCUSSION In the fields of agricultural economics, rural sociology, and agricultural geography, much work has been done investigating the human factors associated with the decision making process, the adoption of management procedures, and farm performance (Jones, 1967 ; Muggen, 1969 ; Ilbery, 1978) . In veterinary medicine, however, although some authors have implicated the farm manager as a contributor to the variation in farm performance, few have actually attempted to measure the relative impact of the human factor . Overall, the RZ of our models were higher than those reported by Bigras-Poulin et
al. (1984-85) . As reported by these authors the variables measuring farmers' personal characteristics appeared regularly in the regression models, indicating that these variables are a basic component of the dairy farm system . With respect to the sociodemographic variables, geographical location was associated with cell counts. This factor has been reported to be a major source of variation on mastitis incidence and prevalence among herds (Bakken, 1982) . However, as in the latter report, our study did not take into consideration many disease determinants that may vary in occurrence among different areas . LEVELIV and HOLIDAY were found to be
1 62
BRITISH VETERINARY JOURNAL, 146, 2
associated with both milk yield and quality . Other studies in Ireland have shown that level of living influences farm performance (Frawley et al. 1974-75a, b) . Although it can be validly argued that level of living and taking holidays are consequences rather than logic antecedents of good farming performance, the demand aspects of the factors can be highly significant (Cepede, 1970 ; Frawley et al., 1974-75a) . Favourable attitudes towards credit and risk, high score in the index measuring information-seeking behaviour, good household structure, and high level of participation in voluntary farm organizations have all been reported to positively influence farm performance in Ireland (Frawley el al., 1974-75b ; Frawley, 1985) . It has been also suggested that the use of credit can be indicative of a progressive attitude towards farm development in general (Frawley, 1985) . Farmers' attitude towards milking was inversely related to somatic cell counts, and positively associated with milk production . Milking practices can play a major role in udder health, and some key mastitis control procedures are applied during milking time . It is conceivable that a positive attitude towards milking might influence farm performance primarily through the correct application of management procedures . Awareness of subclinical mastitis was related to bacterial counts and milk/fat production, but surprisingly, although significantly associated with cell counts at the bivariate level, it did not enter the final models . A farmer aware of subclinical mastitis, however, is more likely to implement control measures that will influence herd cell counts and milk/fat production (Goodhope & Meek, 1980) . Of the set of variables that measured farmers' values, working in family and being independent were associated with low cell counts and high fat production respectively . Continuing the family tradition in farming, however, was inversely related to milk production . It is interesting to note that TRADTION scores had a tendency to increase with increasing herd sizes . A noticeable exclusion from the models was farmer's age . This lack of association was also noticed by Frawley et al. (1974-75a) who warned against the general emphasis age receives in the formulation of some agricultural policies . In accordance with Nygárd (1979), and (sterás & Lund (1988a), a negative regression was found between herd size and mastitis status . Not surprisingly increasing farm and herd sizes were associated with higher milk yields per cow . Of the variables used as proxys of management, dry cow therapy consistently appeared in the regression models, associated with higher milk yields and milk quality . This was expected, since the treatment of dry cows at drying off is known to be linked to low somatic cell counts (Goodhope & Meek, 1980) . Despite the considerable evidence pointing out many individual milking machine factors that predispose to mastitis and high cell counts, some conflicting reports have been published on the epidemiological importance of these factors (Schmidt Madsen & Klastrup, 1980 ; Bakken, 1982 ; Osterás & Lund, 1988b) . In our study, the frequency of *milking machine testing first enters in the SCC300 model, with an inverse relationship . Absence of at least one annual test has already been shown to be a risk factor for mastitis (Barnouin et al., 1986a) . Moreover, old milking plants were associated with high bacterial counts, and type of milking machine was linked to high bacterial counts (bucket plants), and low milk yields (bucket plants and around-the-byre pipelines), agreeing with previous reports (Brandsma & Maatze, 1980 ; Goodhope & Meek, 1980 ; Rabold & Pichler, 1980) . Udder washing can influence milk quality . Udder surfaces and teats should be clean and dry at machine attachment (Galton et al., 1982), since moisture laden with bacteria can drain into the teat cups during milking (Galton, Peterson & Merrill, 1984) . Our study showed improper udder washing and drying to be associated with higher milk cell
FARMERS' PERSONAL CHARACTERISTICS AND FARM PERFORMANCE
1 63
counts and lower milk production . Breeding their own heifers for replacements and high culling were both associated with low cell counts, as reported by Goodhope & Meek (1980) and Vagsholm, Carpenter & jasper (1988) respectively . Type of bedding has been reported to be associated with bacterial population (Rendos, Eberhart & Kesler, 1975), and milk cell counts (Osterás & Lund, 1988a) . However, bacterial counts under barn conditions are influenced by factors more complex than type of bedding alone (Zehner et al., 1986) and may include several other housing characteristics (Saloniemi, 1980 ; Bakken, 1982 ; Barnouin et al., 1986b ; (sterás & Lund, 1988a) . In our study, housing and bedding types were linked with both milk production and milk quality . A difficulty with multiple regression analysis is the problem of collinearity or multicollinearity . This arises when one regressor is nearly a linear combination of other regressors in the model . While the measurement procedures for the explanatory variables are independent, it is often the case that the independent variables themselves are intercorrelated and this can lead to unstable results . In our study collinearity diagnostics showed no such problems . With regard to the variables that entered the regression models, no causal relationships between independent and dependent variables should be implied . These are observed and not causal associations . Moreover, the independent variables used on this study are not to be taken as sufficient or necessary conditions to farm success in terms of good milk/fat production and low somatic cell and bacterial counts . Multiple regression was used, not to construct predictive models, but rather to describe the factors affecting those dependent variables . The aim was to assess the relative importance of the associations between human or management variables with farm performance as measured by the R 2 . Although using a different set of variables to measure both farm performance and the factors that influence it, our results are similar to those reported by Bigras-Poulin et al. (1984-85), and may indicate that those results were not time and/or sample related . In all models but one (XGRADE), the group of variables related to the farmer's attitudes, values, and sociodemographic profile explained a similar or greater amount (between 14 . 44 and 34 . 35%) of the variation of farm performance than the group of management variables (between 14 . 33 and 25 . 9%) . This explanatory capacity of the human variables could explain why, after many years when well proved management procedures to improve both milk yield and milk quality have been available, there is still a large variation among farm performances . In the population of farms under study, there are still plenty of targets to achieve, as the average of almost 6 months per herd per year of herd somatic cell counts over 300 000 suggests . Although this exploratory type of study tends to raise more questions than actually provide definite answers (Frawley, 1985), it stresses the importance of the farmer's personal and social characteristics and the need of addressing them more properly to achieve a wider and more appropriate use of available techniques to improve milk yield and quality .
ACKNOWLEDGEMENTS To all the farmers involved in this study . To Premier Tir Laighean Group, Dublin District Milk Board, Dr J . Frawley, Professors J . Hannan and D . Collins . The senior author was supported by INTA (Argentina) .
164
BRITISH VETERINARY JOURNAL, 146, 2 REFERENCES
AN FORAS TALUNTAIS (1984). Demographic, Structural and Development Aspects of Farm Management . Rural Sociology Department. BAKKEN, G . (1982) . Acta Agriculturae Scandinavica 32, 23 . BARNOUIN, J ., FAYET, J . C ., JAY, M . & BROCH .ART, M . (1986a). Canadian Veterinary Journal27, 135 . BARNOUIN, J ., FAYET, J . C ., JAY, M ., BROCHART, M . & FAYE, B . (1986b) . Canadian Veterinary Journal 27, 173 . BIGRAS-POULIN, M ., MEEK, A. H . & MARTIN, S . W . (1984-85) . Preventive Veterinary Medicine 3, 241 . BRANDSMA, S . & MAATZE, K. (1980) . Proceedings International Workshop Milking Machine and Mastitis . Moorepark Research Centre, Fermoy, Co . Cork, Ireland, 137 . CEPEDE, M . (1970) . Sociologia Ruralis 10, 361 . FRANCIS, P . G . (1984). Proceedings XIIth World Congress on Diseases of Cattle . World Buiatrics Association, Durban, South Africa . Sept. 17-21, 1984, 1, 193 . FRAWLEY, J . (1985) . Proceedings Economy & Rural Welfare Research Centre 12th Annual Conference . Dublin, Ireland, Dec . 3, 1985, 31 . FRAWLEY, J ., BOHLEN, J . M . & BREATHNACH, T . (1974-75a) . Irish Journal of Agricultural Economy and Rural Sociology 5, 145 . FRAWLEY, J ., BOHLEN, J . M . & BREATHNACH, T. (1974-75b) . Irish Journal of Agricultural Economy and Rural Sociology 5, 157 . GALTON, D. M ., ADKINSON, R. W ., THOMAS, C . V . & SMITH, T . W . (1982). Journal of Dairy Science 65, 1540 . GALTON, D. M ., PETERSON, L . G . & MERRILL, W . G . (1984) . Proceedings 23rd Annual Meeting National Mastitis Council. Kansas City, MO, USA, 19 . GASSON, R. (1973) . Journal of Agricultural Economy 24, 521 . GOODHOPE, R . G . & MEEK, A . H . (1980) . Canadian Journal Comparative Medicine 44, 351 . ILBERY, B . W . (1977) . T jdschrift voor Economische en Sociale Geografie 68, 66 . ILBERY, B . W . (1978) . Progress in Human Geography 2, 448 . ILBERY, B . W . (1983) . Transactions. Institute of British Geographers 8, 329 . JONES, G . W . (1967) . World Agricultural Economy and Rural Sociology Abstracts 9, 1 . JONES, G . E . & DAW, M . E . (1964) . Farm Management Notes. University of Nottingham, Sutton Bonington 32, 20. KLEINBAUM, D . G . & KOPPER, L . L . (1982) . Applied Regression Analysis and other Multivariable Methods. Massachusetts : Duxbury Press. MUGGEN, G . (1969). World Agricultural Economy and Rural Sociology Abstracts 11, 1 . NYGARD, A . (1979) . Agricultural University of Norway, Report No 67, 58 . OSTERAS, O . & LUND, A . (1988a) . Preventive Veterinary Medicine 6, 79 . OSTERAS, O . & LUND, A . (1988b) . Preventive Veterinary Medicine 6, 91 . RABOLD, K. & PICHLER, O . (1980) . Proceedings International Workshop Milking Machine and Mastitis . Moorepark Research Centre, Fermoy, Co . Cork, Ireland, 121 . RENDOS, J . J ., EBERHART, R . J . & KESLER, E . M . (1975) . Journal of Dairy Science 58, 1492 . SALONIEMI, H . (1980) . Journal of the Scientific Agricultural Society of Finland 52, 85 . SCHMIDT MADSEN & KLASTRUP, O . (1980) . Proceedings International Workshop Milking Machine and Mastitis . Moorepark Research Centre, Fermoy, Co . Cork, Ireland, 137 . SAS INSTITUTE INC . (1985a) . SAS User's Guide: Basics, version 5 edition . Cary, NC:SAS Institute Inc . SAS INSTITUTE INC . (1985b) . SAS User's Guide: Statistics, version 5 edition . Cary, NC :SAS Institute Inc . VAGSHOLM, J ., CARPENTER, T. E . & JASPER, D . E . (1988) . Preventive Veterinary Medicine 6, 223 . ZEHNER, M . M ., FARNSWORTH, R. J ., APPLEMAN, R. D ., LARNTZ, K . & SPRINGER, J . A . (1986) . Journal of Dairy Science 69, 1932 . (Accepted for publication 27 June 1989)
ASSOCIATIONS BETWEEN FARMERS' PERSONAL CHARACTERISTICS, MANAGEMENT PRACTICES AND FARM PERFORMANCE
H . D . TARABLA and K. DODD Faculty of Veterinary Medicine, Department of Farm Animal Clinical Studies, University College Dublin, Ballsbridge, Dublin 4, Ireland
SUMMARY A survey was carried out in a random sample of 123 dairy farms from the east of Ireland . The monthly mean production per cow was 315 1 of milk and 11 . 5 kg of fat . The mean log herd somatic cell count was 5 . 45 (arithmetic mean=372 573 cells/ml), with almost 50% of the monthly counts over 300 000 cells/ml in a 12month period . Bivariate and multivariate analysis was performed to assess the relative impact of the personal characteristics of the farmer and the management policies he applied on the amount and quality of the milk produced . In five out of six models the group of variables related to farmers' attitudes, values, and sociodemographic profile explained a similar or greater amount (between 14 . 44 and 34 . 35%) of the variation of farm performance than the group of management variables (between 14 . 33 and 25 . 99%) as measured by the R 2 . These results stress the importance of the human factors in explaining variation in farm performance .
INTRODUCTION Only part of the variation in farm performance is due to differences in the quality of land, scale of operation, labour and capital inputs . The rest of the variation must be explained mainly by the factor management and the human factors linked to the adoption of management procedures (Muggen, 1969) . Conventional farm management studies have largely ignored the latter, showing no concern for the fact that the enormous growth in agricultural technology must be ultimately incorporated by individual farmers (Frawley, Bohlen & Breathnach, 1974-75a) . They are by no means an homogeneous sector of society (Gasson, 1973), and their personal and social profile affects the choice of techniques to be applied (Jones & Daw, 1964) and farm performance (Frawley, Bohlen & Breathnach, 1974-75a, b ; Bigras-Poulin, Meek & Martin, 1984-85 ; Frawley, 1985) . As far as mastitis is concerned, despite the fact that farmer's attitudes, knowledge and awareness have long had scope for change or improvement (Francis, 1984), studies have largely dealt with the effect of the environmental and management factors on the disease . The objective of this study was to assess the relative impact of the personal characteristics of the farmer and the management practices he applied on the amount and quality of the milk being produced .
158
BRITISH VETERINARY JOURNAL, 146, 2 MATERIALS AND METHODS
A survey was conducted during the summer of 1987 in 123 randomly selected dairy farms from the east of Ireland . Sociodemographic and psychological variables and management variables were collected using a structured questionnaire . In all cases the interviewee was the farm manager and the interviewer the senior author . One week before the interview a presentation letter was posted to the interviewee . Whenever possible, telephone appointments were made afterwards . Each farmer was paid at least five visits on three different days the aim being to reach all farmers selected . Only two farmers refused to be interviewed and had to be replaced . The questionnaire dealt with a wider spectrum than that covered in this paper, and was tested and adjusted three times before the actual survey . Validity and reliability of the answers were dealt with as suggested by Bigras-Poulin et al. (1984-85) . A summary of the variables used in this study is provided in Table I . To measure awareness of subclinical mastitis (AWARE), a simple definition of subclinical mastitis was provided and the farmers were asked to mention ways of detecting this subclinical stage at the cow or herd level . A correct answer was taken as a proxy for mastitis awareness . The research procedures to assess goals and values such as TRADTION, FAMILY, INDEP, PROFIT, and JOB had been developed by Ilbery (1977, 1983), using approaches suggested by Gasson (1973) . HOUSHLD, INFSEEKB, LEVELIV, RISK, and CREDIT were measured using the approaches developed during previous research in Ireland (An Foras Taluntais, 1984 ; Frawley, 1985) . Farm performance as regards mastitis status, milk yield, and quality was measured for the previous 12 months by the following variables : mean log herd somatic cell count (XLOGSCC), proportion of months with SCC > 300 000 (SCC300), mean grade based on total bacterial count (XGRADE), mean milk yield per cow (XCOWYLD), mean milk yield per cow in the month of peak production (PKCOWYLD), and mean fat production per cow (XFATPROD) . Descriptive statistics were generated for all variables . Normality was checked using normal probability plots, and the Kolmogorov-Smirnov D statistic was used to test goodness of fit. To look for associations between farmer's personal characteristics, management practices, and farm performance, analysis of the data was carried out in three stages, and was a modification of the approach used by Bigras-Poulin et al. (198485) . First, bivariate tests for individual associations between the independent variables on each of the dependent variables were performed (analysis of variance and Student t-test for categorical data and correlation coefficients for continuous data) . Only independent variables that were significantly associated (P<0-15) with each dependent variable in this bivariate screening were subsequently used in the multivariate analysis . Scatter diagrams between continuous and dependent variables were also produced . Monthly herd somatic cell counts were transformed to log s and then averaged to obtain the mean log SCC for each herd . Log transformation was used to avoid the problems of nonhomoscedasticity and positive skewness of the somatic cell count distribution . Second, with the independent variable set relevant for each dependent variable, six separate multiple linear stepwise regression analyses were performed . Graphical analysis of residuals and colinearity diagnostics were also carried out . Cook's D statistic was calculated to detect influential observations . Nominal variables were included in the regression equation as k-1 dummy variables to index the k categories of the variable of interest (Kleinbaum & Kupper, 1982) . Third, the R 2 was calculated for a model
cgothv.Aernwmypuasfl
FARMERS' PERSONAL CHARACTERISTICS AND FARM PERFORMANCE
159
(SAS Institute Inc ., 1985a, b) .
Table I Farmer's personal characteristics and management practices under study in 123 dairy farms from the east of Ireland, 1987
Farmer's profile
Management practices
Sociodemographic AGE Age of farm manager EDUCAT Number of years of formal education DEPEND Number of farm dependents HOUSHLD Household structure INFSEEKB Information seeking behaviour
TOTMEET
COUNTY HATOT
HERDSIZE LEVELIV HOLIDAY
EMPLOYEE
Number of farmer's associations meetings attended last 12 months Geographical location Workable land (ha) (owned + rented) Milking Bows + milking heifers + dry cows Level of living Holidays last 12 months Part or full-time
Values and attitudes TRADTION Continuing farming tradition FAMILY Working close to family INDEP Independence PROFIT Maximize income JOB Doing worthwhile job SATIFARM Satisfaction with farming AWARE Subclinical mastitis awareness CONTVIEW Farmer's view of mastitis control LIKEMILK Attitude towards milking INNOV Innovativeness RISK Risk-willingness Use of credit CREDIT
DAIRY
PROPDRY
HEIFMILK
SORCREP
HOUSNG BEDDING SLUROFTN M-IMTYPE MMOLD
TESTOFFN
BULKTANK MILKOFTN STRIP WUDER
WUDERHW HOTWATER DRYCWTH TEATDIP CULL RECORDS IDENTCOW
MILKLAST
WDRWALL
WDRWDAYS
Workable land used in dairying (%) Proportion of dry cows(%) Proportion of first calved heifers in milk (%) Source of replacements
Winter housing type Bedding type Slurry scraping (times/week) Milking machine type Number of years of milking machine Milking machine testing (times/year) Milk cooling system How often the farmer milks (times/week) Foremilk stripping Udder washing (proportion of cows) Udder washing (how) Hot water where the cows are milked Dry cow therapy Teat dip/spray Culling (%) Mastitis records Identification of cows under mastitis treatment Cows with mastitis are separately milked Withdraws milk from all 4 quarters after antibiotic treatment Withdrawal period (days)
1 60
BRITISH VETERINARY JOURNAL, 146, 2 RESULTS
On average, almost half the monthly cell counts were over 300 000 cells/ml, and that figure has to be related to an overall arithmetic herd somatic cell count mean of 372 573 cells/ml (mean log=5 . 45) . SCC300 and XGRADE had different degrees of departure from normality . XGRADE was positively skewed and leptokurtic (had a high peak), meaning that most milk samples were close to the lower values of the scale (0 being premier quality) . Conversely, SCC300 showed very little skewness and was platikurtic (the distribution was relatively flat and with short tails) . The average cow produced approximately 315 1 of milk and 11 . 5 kg of fat per month, while the average cow during the month of herd peak production yielded 3921 of milk. After multiple stepwise regression human variables appeared in all regression models (Table II) .
Table II Multiple stepwise regression : herd milk quality and production functions (variables and P values), Ireland, 1987 XLOGSCC
SCC300 XGRADE XCOWYLD
PKCOWYLD
XFATPROD
DRYCWTH (0 . 001), LIKEMILK (0 . 02), COUNTY (0 . 02), FAMILY (0 . 03),
SORCREP1 (0 -l), WUDER2 (0 . 1) TESTOFTN (0 . 02), LIKEMILK (0 . 03), RISK1 (0 . 01), HERDSIZE (0- l), CULL (0 . 06), BEDDINGI (0 . 07), BEDDING5 (0 . 1) HOUSNG1 (0 . 0001), AWARE (0 . 0001), DRYCWTH (0 . 002), MMTYPEI (0 . 003), WDRWDAYS (0 . 06), LEVELIV (0 . 09), MMOLD (0 . 1), INFSEEKB (0 . 1), HOLIDAY (0 . 1) DRYCWTH (0 . 0001), AWARE (0 . 0009), MMTYPEI (0 . 006), WUDERHWI (0 . 01), TRADTION (0 . 009), HOLIDAY (0 . 02), MMTYPE2 (0 . 05), INFSEEKB (0 . 02), HOTWATER (0 . 06), WDRWALL (0 . 1), TOTMEET (0 . 1) DRYCWTH (0 . 0001), TRADTION (0 . 02), INNOV (0 . 03), WDRWALL (0 . 06), WUDERHWI (0 . 05), HERDSIZE (0 . 09), CULL (0 . 003), LIKEMILK (0 . 09), TOTMEET (0 . 01), HATOT (0- l), MMTYPEl (0- l), INFSEEKB (0 -ij, HOLIDAY (0 . 1) AWARE (0 . 000l), DRYCW'TH (0 . 003), CREDIT (0 . 02), INDEP (0 . 1)
Low log cell counts were associated with geographical location, the use of dry cow therapy, breeding its own replacements, a positive attitude towards milking and working in family . High log cell counts were associated with washing some cows only . Having a high proportion of monthly counts over 300 000 cells was associated with small herds, irregular testing of the milking machine, type of bedding, and low risk-willingness, while low SCC300 counts were linked to a high score in the farmer's attitude towards milking and high culling . Low total bacterial counts were associated with the use of dry cow therapy and high counts with `tide up' housing, old milking machines, bucket plants, shorter withdrawal periods after antibiotic treatment, and low score in information seeking behaviour . High milk yields were linked with bigger herds, the use of dry cow therapy, good information seeking behaviour, high awareness of subclinical mastitis and high culling . Meanwhile, low yields were associated with bucket plants and barn
FARMERS' PERSONAL CHARACTERISTICS AND FARM PERFORMANCE
161
pipelines, lack of hot water where the cows are milked, the use of a single wet cloth to wash the udder, low attendance at farming related meetings, high score in continuing the family farming tradition and no holidays . High fat yields were associated with treating cows at drying off, the use of credit, high independence, and high awareness of subclinical mastitis. Residual analysis did not show any violation of multiple linear regression analysis assumptions. No significant influential observations or collinearity problems were detected . The models explained between 24 . 45 and 66 . 40% of the variation of the dependent variables . The variables addressing farmers' personal characteristics explained as much or more variation of XLOGSCC, SCC300, XCOWYLD, PKCOWYLD, and XFATPROD than management variables (Table III) .
Table III Contribution to the variation on herd milk yield and quality made by human and management variables (R Z ) in 123 herds, Ireland, 1987
Variables Dependent
XLOGSCC SCC300 XGRADE XCOWYLD PKCOWYLD XFATPROD
Model Independent
Personal
Management
14 . 40 15 . 99 25 . 86 34 . 35 30 . 80 24 . 71
14 . 33 11 . 41 42 . 43 18 . 70 25 . 99 11 . 31
24 . 45 27 . 23 66 . 40 50 . 79 47 . 85 39 . 60
DISCUSSION In the fields of agricultural economics, rural sociology, and agricultural geography, much work has been done investigating the human factors associated with the decision making process, the adoption of management procedures, and farm performance (Jones, 1967 ; Muggen, 1969 ; Ilbery, 1978) . In veterinary medicine, however, although some authors have implicated the farm manager as a contributor to the variation in farm performance, few have actually attempted to measure the relative impact of the human factor . Overall, the RZ of our models were higher than those reported by Bigras-Poulin et
al. (1984-85) . As reported by these authors the variables measuring farmers' personal characteristics appeared regularly in the regression models, indicating that these variables are a basic component of the dairy farm system . With respect to the sociodemographic variables, geographical location was associated with cell counts. This factor has been reported to be a major source of variation on mastitis incidence and prevalence among herds (Bakken, 1982) . However, as in the latter report, our study did not take into consideration many disease determinants that may vary in occurrence among different areas . LEVELIV and HOLIDAY were found to be
1 62
BRITISH VETERINARY JOURNAL, 146, 2
associated with both milk yield and quality . Other studies in Ireland have shown that level of living influences farm performance (Frawley et al. 1974-75a, b) . Although it can be validly argued that level of living and taking holidays are consequences rather than logic antecedents of good farming performance, the demand aspects of the factors can be highly significant (Cepede, 1970 ; Frawley et al., 1974-75a) . Favourable attitudes towards credit and risk, high score in the index measuring information-seeking behaviour, good household structure, and high level of participation in voluntary farm organizations have all been reported to positively influence farm performance in Ireland (Frawley el al., 1974-75b ; Frawley, 1985) . It has been also suggested that the use of credit can be indicative of a progressive attitude towards farm development in general (Frawley, 1985) . Farmers' attitude towards milking was inversely related to somatic cell counts, and positively associated with milk production . Milking practices can play a major role in udder health, and some key mastitis control procedures are applied during milking time . It is conceivable that a positive attitude towards milking might influence farm performance primarily through the correct application of management procedures . Awareness of subclinical mastitis was related to bacterial counts and milk/fat production, but surprisingly, although significantly associated with cell counts at the bivariate level, it did not enter the final models . A farmer aware of subclinical mastitis, however, is more likely to implement control measures that will influence herd cell counts and milk/fat production (Goodhope & Meek, 1980) . Of the set of variables that measured farmers' values, working in family and being independent were associated with low cell counts and high fat production respectively . Continuing the family tradition in farming, however, was inversely related to milk production . It is interesting to note that TRADTION scores had a tendency to increase with increasing herd sizes . A noticeable exclusion from the models was farmer's age . This lack of association was also noticed by Frawley et al. (1974-75a) who warned against the general emphasis age receives in the formulation of some agricultural policies . In accordance with Nygárd (1979), and (sterás & Lund (1988a), a negative regression was found between herd size and mastitis status . Not surprisingly increasing farm and herd sizes were associated with higher milk yields per cow . Of the variables used as proxys of management, dry cow therapy consistently appeared in the regression models, associated with higher milk yields and milk quality . This was expected, since the treatment of dry cows at drying off is known to be linked to low somatic cell counts (Goodhope & Meek, 1980) . Despite the considerable evidence pointing out many individual milking machine factors that predispose to mastitis and high cell counts, some conflicting reports have been published on the epidemiological importance of these factors (Schmidt Madsen & Klastrup, 1980 ; Bakken, 1982 ; Osterás & Lund, 1988b) . In our study, the frequency of *milking machine testing first enters in the SCC300 model, with an inverse relationship . Absence of at least one annual test has already been shown to be a risk factor for mastitis (Barnouin et al., 1986a) . Moreover, old milking plants were associated with high bacterial counts, and type of milking machine was linked to high bacterial counts (bucket plants), and low milk yields (bucket plants and around-the-byre pipelines), agreeing with previous reports (Brandsma & Maatze, 1980 ; Goodhope & Meek, 1980 ; Rabold & Pichler, 1980) . Udder washing can influence milk quality . Udder surfaces and teats should be clean and dry at machine attachment (Galton et al., 1982), since moisture laden with bacteria can drain into the teat cups during milking (Galton, Peterson & Merrill, 1984) . Our study showed improper udder washing and drying to be associated with higher milk cell
FARMERS' PERSONAL CHARACTERISTICS AND FARM PERFORMANCE
1 63
counts and lower milk production . Breeding their own heifers for replacements and high culling were both associated with low cell counts, as reported by Goodhope & Meek (1980) and Vagsholm, Carpenter & jasper (1988) respectively . Type of bedding has been reported to be associated with bacterial population (Rendos, Eberhart & Kesler, 1975), and milk cell counts (Osterás & Lund, 1988a) . However, bacterial counts under barn conditions are influenced by factors more complex than type of bedding alone (Zehner et al., 1986) and may include several other housing characteristics (Saloniemi, 1980 ; Bakken, 1982 ; Barnouin et al., 1986b ; (sterás & Lund, 1988a) . In our study, housing and bedding types were linked with both milk production and milk quality . A difficulty with multiple regression analysis is the problem of collinearity or multicollinearity . This arises when one regressor is nearly a linear combination of other regressors in the model . While the measurement procedures for the explanatory variables are independent, it is often the case that the independent variables themselves are intercorrelated and this can lead to unstable results . In our study collinearity diagnostics showed no such problems . With regard to the variables that entered the regression models, no causal relationships between independent and dependent variables should be implied . These are observed and not causal associations . Moreover, the independent variables used on this study are not to be taken as sufficient or necessary conditions to farm success in terms of good milk/fat production and low somatic cell and bacterial counts . Multiple regression was used, not to construct predictive models, but rather to describe the factors affecting those dependent variables . The aim was to assess the relative importance of the associations between human or management variables with farm performance as measured by the R 2 . Although using a different set of variables to measure both farm performance and the factors that influence it, our results are similar to those reported by Bigras-Poulin et al. (1984-85), and may indicate that those results were not time and/or sample related . In all models but one (XGRADE), the group of variables related to the farmer's attitudes, values, and sociodemographic profile explained a similar or greater amount (between 14 . 44 and 34 . 35%) of the variation of farm performance than the group of management variables (between 14 . 33 and 25 . 9%) . This explanatory capacity of the human variables could explain why, after many years when well proved management procedures to improve both milk yield and milk quality have been available, there is still a large variation among farm performances . In the population of farms under study, there are still plenty of targets to achieve, as the average of almost 6 months per herd per year of herd somatic cell counts over 300 000 suggests . Although this exploratory type of study tends to raise more questions than actually provide definite answers (Frawley, 1985), it stresses the importance of the farmer's personal and social characteristics and the need of addressing them more properly to achieve a wider and more appropriate use of available techniques to improve milk yield and quality .
ACKNOWLEDGEMENTS To all the farmers involved in this study . To Premier Tir Laighean Group, Dublin District Milk Board, Dr J . Frawley, Professors J . Hannan and D . Collins . The senior author was supported by INTA (Argentina) .
164
BRITISH VETERINARY JOURNAL, 146, 2 REFERENCES
AN FORAS TALUNTAIS (1984). Demographic, Structural and Development Aspects of Farm Management . Rural Sociology Department. BAKKEN, G . (1982) . Acta Agriculturae Scandinavica 32, 23 . BARNOUIN, J ., FAYET, J . C ., JAY, M . & BROCH .ART, M . (1986a). Canadian Veterinary Journal27, 135 . BARNOUIN, J ., FAYET, J . C ., JAY, M ., BROCHART, M . & FAYE, B . (1986b) . Canadian Veterinary Journal 27, 173 . BIGRAS-POULIN, M ., MEEK, A. H . & MARTIN, S . W . (1984-85) . Preventive Veterinary Medicine 3, 241 . BRANDSMA, S . & MAATZE, K. (1980) . Proceedings International Workshop Milking Machine and Mastitis . Moorepark Research Centre, Fermoy, Co . Cork, Ireland, 137 . CEPEDE, M . (1970) . Sociologia Ruralis 10, 361 . FRANCIS, P . G . (1984). Proceedings XIIth World Congress on Diseases of Cattle . World Buiatrics Association, Durban, South Africa . Sept. 17-21, 1984, 1, 193 . FRAWLEY, J . (1985) . Proceedings Economy & Rural Welfare Research Centre 12th Annual Conference . Dublin, Ireland, Dec . 3, 1985, 31 . FRAWLEY, J ., BOHLEN, J . M . & BREATHNACH, T . (1974-75a) . Irish Journal of Agricultural Economy and Rural Sociology 5, 145 . FRAWLEY, J ., BOHLEN, J . M . & BREATHNACH, T. (1974-75b) . Irish Journal of Agricultural Economy and Rural Sociology 5, 157 . GALTON, D. M ., ADKINSON, R. W ., THOMAS, C . V . & SMITH, T . W . (1982). Journal of Dairy Science 65, 1540 . GALTON, D. M ., PETERSON, L . G . & MERRILL, W . G . (1984) . Proceedings 23rd Annual Meeting National Mastitis Council. Kansas City, MO, USA, 19 . GASSON, R. (1973) . Journal of Agricultural Economy 24, 521 . GOODHOPE, R . G . & MEEK, A . H . (1980) . Canadian Journal Comparative Medicine 44, 351 . ILBERY, B . W . (1977) . T jdschrift voor Economische en Sociale Geografie 68, 66 . ILBERY, B . W . (1978) . Progress in Human Geography 2, 448 . ILBERY, B . W . (1983) . Transactions. Institute of British Geographers 8, 329 . JONES, G . W . (1967) . World Agricultural Economy and Rural Sociology Abstracts 9, 1 . JONES, G . E . & DAW, M . E . (1964) . Farm Management Notes. University of Nottingham, Sutton Bonington 32, 20. KLEINBAUM, D . G . & KOPPER, L . L . (1982) . Applied Regression Analysis and other Multivariable Methods. Massachusetts : Duxbury Press. MUGGEN, G . (1969). World Agricultural Economy and Rural Sociology Abstracts 11, 1 . NYGARD, A . (1979) . Agricultural University of Norway, Report No 67, 58 . OSTERAS, O . & LUND, A . (1988a) . Preventive Veterinary Medicine 6, 79 . OSTERAS, O . & LUND, A . (1988b) . Preventive Veterinary Medicine 6, 91 . RABOLD, K. & PICHLER, O . (1980) . Proceedings International Workshop Milking Machine and Mastitis . Moorepark Research Centre, Fermoy, Co . Cork, Ireland, 121 . RENDOS, J . J ., EBERHART, R . J . & KESLER, E . M . (1975) . Journal of Dairy Science 58, 1492 . SALONIEMI, H . (1980) . Journal of the Scientific Agricultural Society of Finland 52, 85 . SCHMIDT MADSEN & KLASTRUP, O . (1980) . Proceedings International Workshop Milking Machine and Mastitis . Moorepark Research Centre, Fermoy, Co . Cork, Ireland, 137 . SAS INSTITUTE INC . (1985a) . SAS User's Guide: Basics, version 5 edition . Cary, NC:SAS Institute Inc . SAS INSTITUTE INC . (1985b) . SAS User's Guide: Statistics, version 5 edition . Cary, NC :SAS Institute Inc . VAGSHOLM, J ., CARPENTER, T. E . & JASPER, D . E . (1988) . Preventive Veterinary Medicine 6, 223 . ZEHNER, M . M ., FARNSWORTH, R. J ., APPLEMAN, R. D ., LARNTZ, K . & SPRINGER, J . A . (1986) . Journal of Dairy Science 69, 1932 . (Accepted for publication 27 June 1989)