Reproductive management of dairy herds in New Zealand: Attitudes, priorities and constraints perceived by farmers managing seasonal-calving, pasture-based herds in four regions

Abstract

AIMS: To examine attitudes, priorities, and constraints pertaining to herd reproductive management perceived by farmers managing seasonal-calving, pasture-based dairy herds in four regions of New Zealand, and to explore how these varied with demographic and biophysical factors.

METHODS: Key decision makers (KDM) on 133 dairy herds in four dairy regions (Waikato, Taranaki, and north and south Canterbury) were interviewed between May and July 2009. They were asked to provide demographic and biophysical data about the farm, and to rate their attitude in relation to their own personality traits, management issues and priorities, and likely constraints affecting reproductive performance in their herds. Associations between demographic factors and attitudes, priorities and constraints were analysed using univariable and multivariable proportional-odds regression models.

RESULTS: Farms in the regions studied in the South Island were larger, had larger herds and more staff than farms in the regions studied in the North Island. The farms in the South Island were more likely to be owned by a corporation, managed by younger people or people who had more education, and the herds were more likely to be fed a higher percentage of supplementary feed. The majority of KDM rated the current genetics, milksolids performance and reproductive performance of their herds as high or very high, and >70% believed that the reproductive performance had remained the same or improved over the preceding 3 years. Despite this, improving reproductive performance was the most highly rated priority for the next 3 years. The constraints considered most likely to have affected reproductive performance in the last 2 years were anoestrous cows, protracted calving periods, and low body condition scores; those considered least likely were artificial breeding and heat detection. Of the variables examined related to attitudes, priorities and likely constraints, there were significant differences between region for 10/40, and with age and occupation of the KDM for 24/40 and 5/40, respectively (p<0.05).

CONCLUSIONS: The majority of KDM reported the current reproductive performance of their herds to be high or very high, yet rated improving reproductive performance as a very high priority for the next 3 years. Mismatch between perceived and actual performance may result in reduced uptake of extension programmes designed to improve performance, and accurate benchmarking may help increase uptake and engagement. Further work is needed to determine whether the attitudes and perceptions about performance of farmers affect the likelihood of changes in their management behaviour which translate to measurable change in the actual reproductive performance of their herds. The variation in attitude, priorities and perceived constraints among age groups and region indicates that design of extension programmes may need to vary with these demographics.

KEY WORDS:

• Dairy herd
• reproduction
• key decision maker
• attitude
• priority
• constraint
• satisfaction

KDM

=

Key decision maker(s)

Introduction

Reproductive performance of dairy herds in New Zealand has been decreasing over the last 15 years (Harris et al. 2006). This to similar to trends in other countries (Butler 2000; Royal et al. 2000; Lucy 2001). In New Zealand's dairy system, 90% of herds operate a spring-calving, seasonal-supply, pasture-based system, that presents unique challenges to the manager of the operation (Verkerk 2003). A pivotal requirement of this system is to achieve a compact calving pattern of 8–12 weeks, to align nutritional demands of cows with patterns of pasture growth. In an analysis of the national dairy database, Harris et al. (2006) identified a 10% decline in the proportion of cows re-calving within the first 42 days of the subsequent seasonal calving period, from 1990 to 2004. However, there is variation in performance among herds; in another national study, in the top quartile of herds, the mean 6‐week in-calf rate was 78% (percentage of cows pregnant within the first 6 weeks of mating), mean 3‐week submission rate was 90%, and mean conception rate to first artificial breeding was 62%, compared with 57%, 67% and 43% for these variables, respectively, for the bottom quartile of herds (Xu and Burton 2003). In 2006, a national needs-analysis survey of 200 dairy farmers in New Zealand showed that 58% were only ‘slightly satisfied’ or ‘not at all satisfied’ with the level of reproductive performance in their herds, but 57% acknowledged that they “had a lot of control over the reproductive performance of their herd” (Burke et al. 2008).

In the dairy system in New Zealand, the key decision maker (KDM) on the farm may be an owner-occupier, a sharemilker, a farm manager, or the manager of an equity partnership. The KDM has a large influence on farm management and hence on herd reproductive performance. Associations between a farmer's attitudes, management practices and the herd's reproductive performance were demonstrated >25 years ago (Bigraspoulin et al. 1985ab). Subsequent work in the United States of America (Cow-en et al. 1989ab) described reproductive performance, and identified univariable associations between it and management factors. However, that study focussed on perceived constraints and did not explore other attitudes. More recent work from Scandinavia used more complex Q-methodology to evaluate how dairy farmers perceived the value of being involved in herd health management programmes that included reproductive management (Kristensen and Enevoldsen 2008). That study demonstrated that teamwork, animal welfare, and dissemination of knowledge were the highest priorities for farmers. In contrast, providers of the herd health programme in that study perceived that production and financial gain were the drivers for the farmers.

The farm manager influences farm management by choosing which management protocols to adopt and the rigour with which they are implemented; both the adoption and rigour of implementation of a particular protocol are more likely if the manager perceives a clear benefit (Bigraspoulin et al. 1985ab). Understanding the attitudes and motivation of farmers is important when designing and implementing extension programmes and activities aimed at helping them improve their farm's performance (Paine 1993). Decision making on-farm is almost always multifactorial and complex, and the process for making these decisions can be over-simplified unless the drivers are fully understood (Nettle et al. 2010; Kristensen and Jakobsen 2011).

No work has been published to date on attitudes of KDM managing dairy herds in New Zealand pertaining to the reproductive management or performance of their herds. This study aimed to examine attitudes, priorities and constraints pertaining to herd reproductive management perceived by KDM managing seasonal-calving, pasture-based dairy herds in four regions of New Zealand, and to explore how these varied with demographic and biophysical factors. The overarching aim was to identify predictors of attitude potentially important when designing and implementing extension programmes to influence the behaviour of KDM.

Materials and methods

Overview of the study

This study was the first of a series of studies involving herds from four regions across New Zealand, collectively called the National Herd Fertility Study. The four regions were chosen to represent a diverse cross section of the dairy industry in New Zealand, both geographically and demographically. The National Herd Fertility Study was a herd-level randomised controlled trial, with dairy herds from four dairying regions of New Zealand systematically allocated to one of three treatment groups, viz one group was offered a reproductive extension programme and subject to on-farm monitoring, a similar-sized control group was not provided with the reproductive extension programme but was subjected to the same on-farm monitoring, and in a second control group neither the reproductive extension programme nor on-farm monitoring was undertaken. Farms in the first two of these three groups were enrolled in the current study, and results of baseline interviews conducted with the KDM identified on each of these farms are reported.

Enrolment process

Herds were selected from the client base of four veterinary practices based in the Waikato and Taranaki regions in the North Island, and north Canterbury and south Canterbury in the South Island. Herds were self-selected from those nominated by the coordinating veterinarians at each practice, and thought likely to meet the selection criteria of the study. Selection criteria were: being a client of the participating veterinary clinic, having >90% of the herd calving annually between 01 June and 30 November (i.e. seasonal, predominantly spring-calving herds), the KDM was expected to remain on the same site for the subsequent 2 years and was willing to (and considered likely to) follow the protocol for the study. Each farm was assumed to have one KDM. KDM of potentially eligible herds were identified by the coordinating veterinarians, and invited to a regional meeting prior to the commencement of the study, where the design of the study was outlined and data-recording commitments were explained. Herds in which KDM agreed to participate were then selected. Four selected herds were excluded before allocation because there was evidence of poor data collection or absence of on-farm data recording software. Of the remaining 144 KDM from herds allocated to groups included in the current study, 133 agreed to be interviewed, yielding a response rate of 90% (133/148).

Development and administration of the questionnaire

Data were collected using face-to-face structured interviews, using a questionnaire of the attitudes and priorities of the KDM, biophysical data about the farm, and aspects of disease control (see Supplementary Table 1 1). The interview captured data for the preceding 12 months, i.e. June 2008 to May 2009. Questions were developed based, in part, on validated questions used in questionnaires sent to farmers from previous studies. The questionnaire was pilot-tested with KDM on six farms not included in the study population, to assess whether their interpretation of the questions were consistent with the requirements of the study.

Table 1. Summary statistics for biophysical variables from 133 seasonal-calving, pasture-based dairy herds from four regions in New Zealand, enrolled in the National Herd Fertility Study during 2009, presented for all herds in the study, and for herds in the study by region, compared with published industry means^a.

Biophysical variable of herd	All herds			Means for herds by region				P-value^b
	Mean	SD	Range	Waikato	Taranaki	North Canterbury	South Canterbury
Effective area of farm (ha)	164.4	72.6	50–350	126.6 (109)	114.5 (89)	221.5 (243)	195 (239)	<0.001
Size of herd	561	279	151–1,200	397 (313)	345 (237)	808 (723)	699 (743)	<0.001
Cows/ha	3.34	0.48	1.96–5.03	3.13 (2.93)	2.98 (2.68)	3.66 (3.35)	3.60 (3.14)	<0.001
Full-time staff equivalents (FTE)	3.3	1.3	1–7	2.7	2.3	4.4	3.8	<0.001
Cows/FTE	169	51	78–396	148	151	187	189	<0.001
Years dairy farming^c	18.9	10.0	1–50	22.8	19.3	16.3	16.9	0.03
^a Mean figures for regions are included in brackets, and were derived from New Zealand Dairy Statistics 2008–09 (Anonymous 2009)
^b Significance of effect of region as a covariate in linear regression models
^c Years the key decision maker interviewed for each farm had been dairy farming

Interviews were conducted with the KDM on each farm between May and July 2009. The interviews were undertaken by four trained technicians, one in each region, using a paper-based questionnaire. These technicians were provided with training in interview technique by a social scientist, and with detailed documentation about the interview process. The interviews took approximately 60 minutes. Subsequently, data were checked for missing values and ambiguous responses, some calculated fields were generated, e.g. stocking rate based on the number of cows and area of the farm, and KDM were re-contacted by technicians to seek clarification of particular responses if necessary.

Biophysical and demographic variables included region, effective area of the farm (number of hectares utilised for grazing milking cows), size and predominant breed of the herd, number of fulltime equivalents of staff employed in the previous season, farm business structure (family-owned, corporation or equity partnership) and production system (based on the five production systems identified by DairyNZ, and summarised in Table 2), and the occupation, age category (20–29, 30–39, 40–49 and ≥50), number of years dairy farming, and further education of the KDM.

Table 2. Summary statistics for demographic variables for farms and key decision makers (KDM) from 133 seasonal-calving, pasture-based dairy herds from four regions in New Zealand, enrolled in the National Herd Fertility Study during 2009, presented for all herds in the study, and for herds in the study by region.

Variable, and level	No. (%) all herds	Herds by region				P-value^a
		Waikato	Taranaki	North Canterbury	South Canterbury
Farm
Farm business structure						0.02
Family owned farm	104 (78%)	91%	88%	67%	66%
Non-family owned corporate farm	13 (10%)	9%	6%	9%	16%
Equity partnership	16 (12%)	0%	6%	24%	19%
Production system^b						<0.001
1	11 (8%)	11%	21%	0%	0%
2	14 (10%)	11%	24%	0%	6%
3	38 (29%)	49%	18%	24%	22%
4	38 (29%)	26%	15%	30%	44%
5	32 (24%)	3%	21%	46%	28%
Predominant breed						0.60
Friesian	51 (38%)	39%	36%	33%	47%
Friesian x Jersey	70 (53%)	58%	49%	58%	44%
Jersey	12 (9%)	3%	15%	9%	9%
KDM
Occupation						0.01
Owner-occupier	77 (57%)	69%	63%	64%	34%
Higher-order sharemilker^c	25 (19%)	20%	31%	6%	19%
Lower-order sharemilker^c	17 (13%)	6%	6%	12%	28%
Farm manager	4 (3%)	3%	0%	3%	6%
Other (manager of equity partnership)	10 (8%)	3%	0%	16%	13%
Age group (years)						0.02
20–29	12 (9%)	3%	13%	9%	13%
30–39	36 (27%)	17%	22%	21%	50%
40–49	56 (42%)	46%	41%	55%	28%
≥50	28 (20%)	34%	25%	15%	9%
^a Significance of differences between regions derived using χ^2 tests
^b Production systems were defined as: 1 = no imported feed fed to lactating or non-lactating (dry) cows; 2 = 4–14% imported feed fed to lactating cows, and dry cows grazed off-farm; 3 = 10–20% imported feed fed to dry cows and to lactating cows at the end of the season to extend lactation; 4 = 20–30% imported feed fed to dry cows and to lactating cows at both the beginning and end of the season to extend lactation; 5 = 25–30% imported feed used throughout the year
^c Higher-order sharemilkers were those that received >35% of farm income; lower-order sharemilkers were those that received ≤35% of farm income

The order of the questions in the questionnaire was designed to avoid directing the thinking of KDM toward herd reproduction until reproduction-specific questions were asked. Some questions related to specific reproductive management options but these were not asked until after all higher-level views had been explored. The attitudes of the KDM toward farm management, priorities for the farm, sources of farming information, staffing, responsibility for tasks, opportunities to improve reproductive performance, and reproductive constraints were explored. Priorities and constraints were nominated and selected based on established reproductive risk factors in herds from an Australian study (Morton 2004) which represented the most recent work in this field for dairy herds in Australia and New Zealand.

Key decision makers were asked to respond to attitude-rating questions using a Likert-type scale of 1 to 5 (Likert 1932), where 1 indicated very low importance, likelihood or priority, and 5 indicated very high importance, likelihood or priority. For example, one question was “Rate the importance of succession planning on your farm, where 1 is very low importance and 5 is very high importance”. All questions relating to attitude were independent of each other. When rating lists of choices regarding perceived priorities or constraints, the KDM was told that not all choices should be rated the same, but there was no restriction on how they rated each one.

Statistical analysis

All data were entered into a relational database written in structured query language designed for the National Herd Fertility Study, and analysed using R (R Development Core Team 2010; R Foundation for Statistical Computing, Vienna, Austria).

The unit of analysis was the individual herd. Continuous data are described using means, SD and range, while categorical data are presented as percentages. Biophysical and demographic data were described, and univariable associations with region assessed using a linear regression model for continuous data, and χ² tests for categorical data. Distributions of responses by KDM on the Likert-type scales were initially assessed using frequency plots. Associations between responses on the Likert-type scales and each demographic variable were assessed using univariable proportional-odds regression models.

Potential instability in the multi-level modelling caused by high correlation between predictor variables was avoided by selecting only the most plausible predictor variable from a correlated set. Correlation between potential predictor variables was investigated by assessing how much the deviance of a Gaussian or binomial generalised linear mixed model, as appropriate (using one variable to predict another), as a proportion of the deviance of the null model (the same model with the outcome only), was ‘explained’ by the predictor, giving a pseudo-R² with a range of 0 to 1 (where 0 is no relationship, and 1 is complete collinearity). For example, the effective area of the farm and number of full-time staff were both highly correlated with the size of the herd (pseudo-R² = 0.91 and 0.74, respectively), so the size of the herd was used as it was a more plausible indicator of attitudes.

The demographic variables selected from the univariate analysis were then assessed using a backwards stepwise model-building procedure and a multivariable proportional-odds regression model until all remaining predictor variables in the model were significantly associated with the attitude (p<0.05). The OR, calculated by exponentiating the coefficients from the regression models, estimate the odds of giving a particular Likert-type rating or higher (rather than a lower rating) for each group relative to a reference group. For example, the odds of KDM from the Waikato region rating their appetite to learn at or above any particular rating were three times higher than those of KDM from the Taranaki region.

Proportional-odds ordinal-regression models assume that the OR are the same throughout the Likert-type scale (the proportional-odds assumption). To test this assumption, the ordinal likelihood ratios were compared with those from multinomial logistic models with the same covariates. Where a significant difference was identified between these models using likelihood ratio tests, a comparison recommended by Faraway (2006), the proportional-odds assumption was assumed to be violated, and the proportional-odds model was rejected. Results from the multinomial models are not presented.

Results

A total of 133 KDM were interviewed, of whom 35 (26%), 33 (25%), 33 (25%) and 32 (24%) were from the Waikato, Taranaki, north Canterbury and south Canterbury regions, respectively. Data were missing from only four attitudinal variables, constituting 0.02% of all attitudinal data.

Biophysical and demographic data reported by KDM interviewed in the study are presented in Tables 1 and 2. Table 1 also includes the regional averages for effective area of the farm, size of the herd, and cows per hectare taken from the national statistics for 2008/2009 published by the Livestock Improvement Corporation (Hamilton, New Zealand) and the national dairy levy body, DairyNZ (Anonymous 2009). Statistics for the herds in the study enrolled in the south Canterbury region were compared with those for the Waimate region in the national statistics, the most similar region geographically. In three of the regions, Waikato, Taranaki and north Canterbury, the mean size of herds in the study was larger than the regional average. Only in south Canterbury was there a smaller mean size of the herd in the study than the reported regional average.

There were no significant differences in reproductive management intervention choices between regions. Induction of parturition was used by 71% of herds, hormonal treatment of anoes-trous cows during breeding by 79%, and routine examination for endometritis (based on the presence and nature of purulent discharge in the vagina) by 74%. However, the relatively less expensive measures of body condition scoring of cows and weighing of young stock were used by only 34% and 38% of KDM, respectively. When explored further, 80% of the KDM who measured body condition score of cows volunteered that the procedure was normally undertaken by an external person (normally a farm advisor) and that they were unsure how many cattle were scored and what method was employed to score each cow (visual-only, or a hands-on approach). Likewise, the weighing of young stock was undertaken by an external grazier in almost all cases, and again KDM were unsure about the protocols employed. In herds in the South Island, weighing of young stock was used by more KDM than body condition scoring of cows, whereas in the North Island, these procedures were used with similar frequency.

Attitudes

Key decision makers were asked to rate 17 attitudes. Most rated their self motivation for their job, their appetite to learn more, and their attention to detail as 4 or 5 on a scale of 5, where 5 was very high (Figure 1). There was a greater range of attitudes toward the perceived benefit gained from group learning experiences and their perceived ability to manage staff. KDM rated their perceived influence over other farmers as mostly 3.

Figure 1. Distribution of responses by key decision makers managing 133 seasonal-calving, pasture-based dairy herds from four regions in New Zealand, enrolled in the National Herd Fertility Study during 2009, asked to rate their attitude in relation to personality traits, using a Likert-type scale, where 1 = very low and 5 = very high.

Regarding attitudes to the importance of various farm management practices, the importance of farm infrastructure, including tracks, fences, water systems and pastures, was rated highest and was least variable (Figure 2). The importance of timeliness of completing seasonal tasks, data recording, following proven protocols, delegation, and farm vehicles, tools and machinery was all rated as high or very high by the majority of farmers (median score = 4). The importance of succession planning was rated below most other attitudes, across all regions. KDM not employing staff were not asked questions relating to staff management.

Figure 2. Distribution of responses by key decision makers managing 133 seasonal-calving, pasture-based dairy herds from four regions in New Zealand, enrolled in the National Herd Fertility Study during 2009, asked to rate the importance of farm management practices or likelihood of them trying unproven ideas or technologies, using a Likert-type scale, where 1 = very low and 5 = very high.

More than 50% of KDM rated the current genetic quality, milksolids performance and reproductive performance of their herds as high or very high, and there was no significant difference between regions (Figure 3). The majority of KDM thought the reproductive performance of their herd had either stayed the same or improved over the last three seasons. The percentage of KDM who thought the reproductive performance of their herd was the same as (Waikato 18%, Taranaki 72%, north Canterbury 54%, and south Canterbury 46%) or better than (Waikato 67%, Taranaki 21%, north Canterbury 36%, south Canterbury 43%) the average for the district varied amongst regions. The remainder felt that the reproductive performance of their herd was lower than that of the regional average.

Figure 3. Distribution of responses by key decision makers managing 133 seasonal-calving, pasture-based dairy herds from four regions in New Zealand, enrolled in the National Herd Fertility Study during 2009, asked to rate the current genetic quality, milksolids performance and reproductive performance of their herd, using a Likert-type scale, where 1 = very low and 5 = very high.

Priorities

Ratings of the priority attributed to 10 areas of management over the next 3 years by KDM are presented in Figure 4. Collectively, 6/10 areas were rated as high priority (median Likert score = 4 or 5), and three of those as very high priority (median Likert score of each = 5). Overall, reproductive management of the herd was given the highest priority, with 127/133 (95%) KDM rating it as 4 or 5 (high or very high). This was closely followed by pasture management (123/133; 92%), business management of the farm (121/133; 91%), and animal health (117/133; 88%). Only 19/133 (14%) KDM felt that environmental sustainability on their farm and milk quality were a low or very low priority. Responses on management of staff as a priority were provided by the 110 KDM who employed staff. Of these, 87/110 (79%) rated management of staff as a high or very high priority (Lik-ert score = 4 or 5). Overall, the range of responses was narrower for management priorities than for attitudes (above); 50% of responses were ± 1 score for 4/10 management priorities.

Figure 4. Distribution of responses by key decision makers managing 133 seasonal-calving, pasture-based dairy herds from four regions in New Zealand, enrolled in the National Herd Fertility Study during 2009, asked to rate their management priorities over the next 3 years, using a Likert-type scale, where 1 = very low and 5 = very high.

Constraints

The likelihoods of identified risk factors perceived by KDM as constraints to their herd's reproductive performance over the previous two lactations are summarised in Figure 5. The spread of responses was wider than for management attitudes and priorities. Only the high prevalence of anoestrous cows in the herd during the first 3 weeks of mating, a protracted calving period (cows calving 9 weeks after the start of calving), and the body condition score of the herd had a median response of 3 (moderately likely constraint). The median response was <3 for all other constraints, i.e. these were perceived unlikely or very unlikely to be a constraint. Heat detection was considered unlikely to be a constraint to their herd's reproductive performance by 93/133 (70%) KDM, and almost all KDM considered artificial breeding and storage of semen were very unlikely to be constraints. When provided opportunity to suggest other likely constraints, 31/133 (23%) respondents did so; these were: nutrition and diet (8/31; 26%); the weather (8/31; 26%); infectious disease (5/31; 16%); physical characteristics of their farm (2/31; 6%); other farm-level management decisions made (9/31; 29%); and staffing (1/33; 3%). There was a regional trend in responses to this optional question; KDM from southern regions were more likely to consider nutrition and those in the northern regions were more likely to consider weather and management decisions to be constraints.

Figure 5. Distribution of responses by key decision makers managing 133 seasonal-calving, pasture-based dairy herds from four regions in New Zealand, enrolled in the National Herd Fertility Study during 2009, asked to rate the likelihood of factors being a constraint to the reproductive performance of their herd over the previous 2 years, using a Likert-type scale, where 1 = very low and 5 = very high.

Univariable and multivariable analyses

One or more independent demographic variables were associated with all but five of the 40 attitudes, priorities and constraints assessed. There was no association between any of the demographic explanatory variables and the perceived importance of farm infrastructure, management of staff, choosing staff to match job roles, and farm business and herd reproduction as a priority for future management. Three of the remaining models did not meet the proportional-odds assumption, and were rejected. These were for the attention to detail, view on milksolids performance, and perceived importance by the KDM of artificial breeding and semen storage as a constraint on herd reproductive performance. Of the 32 attitudes, priorities and perceived constraints examined that had significant predictors, age group of the KDM was significant-ly associated with 24, region with 10, and the occupation of the KDM with 5. Tables 3, 4 and 5 show the results of the multivariable models of responses for 29 attitudes, priorities and perceived constraints examined, respectively; results of the remaining three are presented here as follows.

Table 3. Odds ratios (and 95% CI) of significant predictors for rating attitudinal and management variables (using a Likert-type scale where 1 = very low and 5 = very high) by key decision makers (KDM) from 133 seasonal-calving, pasture-based dairy herds from four regions in New Zealand, enrolled in the National Herd Fertility Study during 2009.

Attitude	Predictor, and level	OR (95%CI)	P-value^a
Self motivation for job	Herd size (per SD of herd size)^b	1.62 (1.15–2.32)	0.006
Appetite to learn	Region^c		0.007
	Waikato	3.01 (1.21–7.66)
	Taranaki	Ref
	North Canterbury	4.57 (1.78–12.14)
	South Canterbury	3.73 (1.47–9.76)
influence over other farmers	KDM age (years)		0.015
	20–29	Ref
	30–39	1.79 (0.51–6.34)
	40–49	1.76 (0.52–5.92)
	≥50	2.18 (0.58–8.19)
Importance of timeliness of seasonal tasks	Region^c		<0.001
	Waikato	6.11 (2.13–19.63)
	Taranaki	Ref
	North Canterbury	4.33 (1.6–12.57)
	South Canterbury	6.76 (2.25–23.61)
Importance of adherence to proven protocols	KDM occupation		0.008
	Owner-occupier	Ref
	Higher-order sharemilker^d	1.19 (0.49–2.90)
	Lower-order sharemilker^d	3.00 (1.00–9.68)
	Farm manager	7.57 (2.02–37.4)
	KDM age (years)		0.020
	20–29	Ref
	30–39	1.87 (0.51–7.00)
	40–49	2.01 (0.54–7.66)
	≥50	3.48 (0.87–14.45)
Likelihood of trying unproven ideas or technologies	Production system^e		0.008
	1	Ref
	2	3.47 (0.78–16.33)
	3	7.66 (2.02–30.95)
	4	6.15 (1.61–24.93)
	5	11.1 (2.78–47.3)
benefit of group learning with farmers	KDM age (years)		0.008
	20–29	2.04 (0.64–6.70)
	30–39	1.75 (0.78–3.94)
	40–49	Ref
	≥50	2.27 (0.95–5.54)
	Production system^e		0.03
	1	1.83 (0.53–6.39)
	2	1.67 (0.51–5.59)
	3	4.21 (1.73–10.58)
	4	Ref
	5	1.54 (0.61–3.95)
	Herd size (per SD of herd size)^b	1.62 (1.10–2.40)	0.01
Importance of farm machinery	KDM age (years)		<0.001
	20–29	3.27 (1.02–11.17)
	30–39	2.46 (1.11–5.56)
	40–49	Ref
	≥50	1.47 (0.62–3.5)
Importance of data recording	Region^c		0.01
	Waikato	4.48 (1.71–12.16)
	Taranaki	Ref
	North Canterbury	3.62 (1.41–9.63)
	South Canterbury	2.76 (1.02–7.63)
	KDM age (years)		0.02
	20–29	1.71 (0.53–5.84)
	30–39	1.55 (0.67–3.67)
	40–49	Ref
	≥50	1.50 (0.61–3.76)
Importance of staff training	KDM occupation		0.03
	Owner-occupier	1.38 (0.51–3.86)
	Higher-order sharemilker^d	Ref
	Lower-order sharemilker^d	3.40 (0.88–13.96)
	Farm manager	6.10 (1.49–29.02)
Importance of staff suggestions	KDM age (years)		0.03
	20–29	Ref
	30–39	5.40 (1.39–21.84)
	40–49	4.14 (1.14–15.57)
	≥50	5.73 (1.4–24.39)
Importance of delegation	KDM age (years)		0.03
	20–29	3.31 (0.62–18.82)
	30–39	3.40 (0.97–12.37)
	40–49	1.62 (0.6–4.38)
	≥50	Ref
	Years dairy farming^f	1.07 (1.02–1.12)	0.009
	Production system^e		0.04
	1	Ref
	2	4.63 (0.95–23.71)
	3	4.23 (1.13–16.53)
	4	4.25 (1.14–16.44)
	5	8.89 (2.28–36.61)
Importance of succession planning	KDM age (years)		0.003
	20–29	1.14 (0.35–3.92)
	30–39	1.94 (0.91–4.18)
	40–49	Ref
	≥50	1.58 (0.69–3.7)
^a Significance of the likelihood-ratio test indicating that a model including the variable improved the ft of the model significantly, with an alpha (type I) error equivalent to p
^b The odds of rating self motivation at any particular rating or higher increased by 1.62 for every increase in herd size of 279 cows (SD of size of herd in the study, from Table 1)
^c The predictor represented the effect of the four regions examined in the study
^d Higher-order sharemilkers were those that received >35% of farm income; lower-order sharemilkers were those that received ≤35% of farm income
^e Production systems were defined as: 1 = no imported feed fed to lactating or non-lactating (dry) cows; 2 = 4–14% imported feed fed to lactating cows, and dry cows grazed off-farm; 3 = 10–20% imported feed fed to dry cows and to lactating cows at the end of the season to extend lactation; 4 = 20–30% imported feed fed to dry cows and to lactating cows at both the beginning and end of the season to extend lactation; 5 = 25–30% imported feed used throughout the year
^f Odds of reporting any particular rating or higher for every extra year dairy farming
Ref = reference (for all variables, the level with the lowest odds was used as the reference category)

Table 4. Odds ratios (and 95% CI) of significant predictors for increased rating of perceived priorities (using a Likert-type scale, where 1 = very low and 5 = very high) by key decision makers (KDM) from 133 seasonal-calving, pasture-based dairy herds from four regions in New Zealand, enrolled in the National Herd Fertility Study during 2009.

Priority	Predictor, and level	OR (95%CI)	P-value^a
Animal health	Region^b		0.028
	Waikato	Ref
	Taranaki	2.49 (0.9–7.16)
	North Canterbury	1.55 (0.57–4.31)
	South Canterbury	6.68 (1.77–31.14)
	KDM occupation		0.013
	Owner-occupier	Ref
	Higher-order sharemilker^c	1.57 (0.53–5.15)
	Lower-order sharemilker^c	28.63 (3.2–737.95)
	Farm manager	1.70 (0.44–7.54)
	KDM age (years)		0.015
	20–29	Ref
	30–39	8.16 (1.52–48.61)
	40–49	11.42 (2.07–70.7)
	≥50	8.02 (1.38–51.85)
Nutrition management	Region^b		0.001
	Waikato	Ref
	Taranaki	2.85 (1.15–7.25)
	North Canterbury	1.82 (0.72–4.65)
	South Canterbury	8.34 (2.9–25.62)
	KDM age (years)		0.003
	20–29	Ref
	30–39	2.53 (0.74–8.75)
	40–49	1.87 (0.59–5.97)
	≥50	3.61 (0.98–13.64)
Environmental sustainability	Region^b		<0.001
	Waikato	Ref
	Taranaki	4.79 (1.93–12.21)
	North Canterbury	1.22 (0.5–3.01)
	South Canterbury	9.50 (3.5–26.91)
	KDM age (years)		0.016
	20–29	Ref
	30–39	1.29 (0.37–4.61)
	40–49	1.71 (0.51–5.86)
	≥50	1.70 (0.46–6.47)
Herd genetics	Region^b		<0.001
	Waikato	Ref
	Taranaki	3.09 (1.26–7.72)
	North Canterbury	1.01 (0.41–2.47)
	South Canterbury	5.88 (2.21–16.22)
	KDM age (years)		0.004
	20–29	1.26 (0.36–4.44)
	30–39	Ref
	40–49	1.26 (0.55–2.86)
	≥50	1.52 (0.59–3.96)
	Further education^d	0.43 (0.21–0.87)	0.018
Milksolids performance	Region^b		0.004
	Waikato	Ref
	Taranaki	4.10 (1.56–11.17)
	North Canterbury	2.65 (1–7.23)
	South Canterbury	5.84 (2.05–17.53)
	KDM age (years)		<0.001
	20–29	Ref
	30–39	5.59 (1.58–20.66)
	40–49	10.62 (2.72–44.4)
	≥50	26.29 (4.65–166.77)
	Years dairy farming^e	0.92 (0.87–0.97)	0.002
Milk quality	Region		<0.001
	Waikato	1.51 (0.64–3.63)
	Taranaki	9.28 (3.44–26.66)
	North Canterbury	Ref
	South Canterbury	10.78 (3.85–32.34)
	KDM age (years)		<0.001
	20–29	Ref
	30–39	3.33 (0.93–11.93)
	40–49	6.01 (1.7–21.6)
	≥50	4.43 (1.13–17.61)
Pasture management	KDM occupation		0.036
	Owner-occupier	1.54 (0.62–3.73)
	Higher-order sharemilker^c	Ref
	Lower-order sharemilker^c	11.89 (1.97–230.09)
	Farm manager	2.30 (0.55–12.03)
Staff management	KDM occupation		<0.001
	Owner-occupier	Ref
	Higher-order sharemilker^c	2.38 (0.85–7.01)
	Lower-order sharemilker^c	18.03 (4.47–84.67)
	Farm manager	6.98 (1.96–27.83)
	KDM age (years)		0.001
	20–29	Ref
	30–39	1.54 (0.38–6.22)
	40–49	8.13 (1.78–38.42)
	≥50	2.93 (0.63–13.96)
^a Significance of the likelihood-ratio test indicating that a model including the variable improved the ft of the model significantly, with an alpha (type I) error equivalent to p
^b The predictor represented the effect of the four regions examined in the study
^c Higher-order sharemilkers were those that received >35% of farm income; lower-order sharemilkers were those that received ≤35% of farm income
^d The odds of increasing a ranking of a priority if the KDM had undertaken a form of further education
^e Odds of reporting any particular rating or higher for every extra year dairy farming
Ref = reference (for all variables, the level with the lowest odds was used as the reference category)

Table 5. Odds ratios (and 95% CI) of significant predictors for increased rating of the perceived likelihood of constraints affecting herd reproductive performance (using a Likert-type scale, where 1 = very unlikely and 5 = very likely) by key decision makers (KDM) from 133 seasonal-calving, pasture-based dairy herds from four regions in New Zealand, enrolled in the National Herd Fertility Study during 2009.

Factor affecting reproduction	Predictor, and level	OR (95%CI)	P-value^a
Anoestrous cows in the herd
	KDM age (years)		0.018
	20–29	2.49 (0.8–7.93)
	30–39	Ref
	40–49	1.38 (0.66–2.88)
	≥50	1.12 (0.43–2.90)
Low body condition score of herd	KDM age (years)		0.031
	20–29	1.08 (0.32–3.58)
	30–39	1.20 (0.49–2.96)
	40–49	1.41 (0.63–3.18)
	≥50	Ref
Management of bulls	Farm business type		0.031
	Family owned	Ref
	Non-family owned	2.81 (0.99–8.15)
	Equity partnership	2.82 (1.04–7.69)
	KDM age (years)		0.006
	20–29	1.11 (0.31–3.89)
	30–39	Ref
	40–49 ≥50	1.21 (0.56–2.62) 1.92 (0.75–4.99)
Protracted	KDM age (years)		0.002
calving spread	20–29	2.83 (0.76–10.6	9)
	30–39	2.02 (0.81–5.05)
	40–49	2.39 (1.03–5.61)
	≥50	Ref
Endometritis	KDM age (years)		0.003
	20–29	5.39 (1.44–20.84)
	30–39	2.04 (0.70–6.00)
	40–49	3.14 (1.19–8.58)
	≥50	Ref
	Further education^b	3.30 (1.57–7.15)	0.002
	Production system^c		0.042
	1	1.83 (0.45–7.41)
	2	2.57 (0.65–9.87)
	3	3.66 (1.52–9.02)
	4	Ref
	5	1.04 (0.41–2.67)
	Herd size (per SD of herd size)^d	1.93 (1.29–2.92)	0.001
Heat detection	KDM age (years)		0.006
	20–29	2.87 (0.91–9.15)
	30–39	1.26 (0.59–2.70)
	40–49	Ref
	≥50	1.57 (0.66–3.76)
Heifer size at calving	Region^e		0.033
	Waikato	3.57 (1.38–9.5)
	Taranaki	1.18 (0.45–3.14)
	North Canterbury	1.85 (0.74–4.7)
	South Canterbury	Ref
	KDM age (years)		<0.001
	20–29	6.57 (1.83–24.59)
	30–39	3.64 (1.27–10.92)
	40–49	4.03 (1.58–10.82)
	≥50	Ref
Genetic quality of herd	KDM age (years)		0.031
	20–29	1.28 (0.38–4.26)
	30–39	1.55 (0.61–3.98)
	40–49	1.40 (0.59–3.34)
	≥50	Ref
^a Significance of the likelihood-ratio test indicating that a model including the variable improved the ft of the model significantly, with an alpha (type I) error equivalent to p
^b The odds of increasing a ranking of a priority if the KDM had undertaken a form of further education
^c Production systems were defined as: 1 = no imported feed fed to lactating or non-lactating (dry) cows; 2 = 4–14% imported feed fed to lactating cows, and dry cows grazed off-farm; 3 = 10–20% imported feed fed to dry cows and to lactating cows at the end of the season to extend lactation; 4 = 20–30% imported feed fed to dry cows and to lactating cows at both the beginning and end of the season to extend lactation; 5 = 25–30% imported feed used throughout the year
^d The odds of rating self motivation at any particular rating or higher increased by 1.62 for every increase in herd size of 279 cows (SD of size of herd in the study, from Table 1)
^e The predictor represented the effect of the four regions examined in the study

Age group was a predictor for how KDM rated the reproductive performance of their herd (p<0.05). Relative to older age groups, younger (20–29‐year-old) KDM were least likely to rate their herd's reproductive performance at or above any particular rating compared with those 30–39, 40–49, and ≥50 years old (OR 2.2, 95% CI = 0.7–7.5; OR 2.10, 95% CI = 0.7–6.8; and OR 3.2, 95% CI = 0.9–11.6 for these groups relative to 20–29‐year-olds, respectively). KDM aged 30–39 years were more likely (OR 1.4 (95% CI = 0.4–5.2); p<0.05) to rate their herd's change in reproductive performance as improved in the last 3 years than those aged 20– 29 years, and the other age groups were intermediate. Region was more strongly associated than age group with the perception of a KDM of their herd's reproductive performance being better than the average for the district. KDM in the Taranaki region were less likely to rate their herd's reproductive performance as better than the average for the district than those in the Waikato, north Canterbury, or south Canterbury (OR 9.8, 95% CI = 3.5–29.6; OR 2.2, 95% CI = 0.82–6.3; and OR 3.0, 95% CI = 1.1–8.7 for these regions relative to Taranaki, respectively; p<0.05).

Key decision makers aged 40–49 years and those who managed large and/or intensive dairy farms (production system 4 and 5) were least likely to consider group learning important. Group learning was rated most important by KDM ≥50 years old, closely followed by those 20–29 years old, and by those who managed semi-intensive production systems (system 3) and/or smaller herds.

Staff training was considered most important by farm managers and least important by higher-order sharemilkers and owner-occupiers. All age groups >29 years old were more likely to rate the importance of staff suggestions highly than 20–29‐year-olds (p = 0.03).

Younger (20–29 years old) KDM rated the constraints to reproductive performance as more likely than older KDM in 7/8 models. In five of these models, the odds of this younger age group rating constraints at or above any particular rating were twice those for the other age groups.

Discussion

The aim of this study was to examine attitudes, priorities, and constraints pertaining to herd reproductive management perceived by KDM managing seasonal-calving, pasture-based dairy herds in four regions of New Zealand, and to explore how these varied with demographic and biophysical factors. The overarching aim was to identify possible predictors of attitude potentially important when designing and implementing management extension programmes to influence KDM, and to help advisors improve the focus of their advice. This is the first published attempt to formally characterise the attitudes of KDM managing dairy herds in New Zealand, hence there are no comparative data. The interpretation of these results is therefore drawn with development of extension programmes in mind.

The farms included in this study were intended to be representative of those in the regions they were drawn from. However, a number of potential biases should be considered. The most obvious are the enrolment criteria. The requirement for good-quality data to be collected on-farm for the duration of the study resulted in selection bias towards KDM that the local veterinarians perceived to be most capable of complying. This may explain the slightly larger herd sizes and areas of the farms included in the study than reported in regional statistics (Anonymous 2009). The opinions of KDM on the importance of data recording varied amongst region and age category but not by size of herd, suggesting that any bias relating to size of herd was unlikely to have biased the conclusions. It could be suggested that KDM willing to collect data are more likely to remain in business in the future, and hence should be the focus of research.

Other biases could have been introduced by veterinarians excluding potential clients they perceived to be difficult to work with, and variation between interviewers, which was totally confounded with region, as only one interviewer per region was used. The study employed local technicians in each region and provided their training. The degree of familiarity each technician had with the KDM they interviewed and their technical understanding of dairy farming also probably varied, and may have affected responses recorded. To mitigate this, interviewers were alerted to this risk, and all questions were discussed during their training to reduce variation in interpretation of responses by KDM. Regular communication between interviewers and the study's management team during the interviewing period helped reduce variability between regions by addressing issues as they were encountered, across the study. Answers varied on average by 13.4% when farmers were interviewed twice in one year as part of a mastitis study (Schukken et al. 1989). Similar variation was reported in an earlier study of responses to an interview (Horwitz and Yu 1985). In the study by Schukken et al. (1989), variation in responses was an accumulation of actual differences in the farmer's answer, and different interpretations by the interviewers of the same answer given by the farmer. Those authors examined objective management choices and performance indicators, whereas the current study was principally concerned with the views and attitudes of the KDM at the start of the period of the study. The repeatability of responses in the current study may have been lower due to the subjective nature of the topics. The views of the KDM were largely collected using Likert-type scales, which meant there was less opportunity for interviewer technique resulting in bias than there might have been with open-ended questions. However, a limitation of the Likert-type scale is acquiescence bias, a tendency of respondents to agree with the statements and gravitate toward the optimal end of the scale (Likert 1932). The design of questions and training of interviewers attempted to minimise this issue in the study reported here.

More than half the KDM rated their herd's reproductive performance as high or very high; none rated it as very low, and 75% considered their herd's reproductive performance had either stayed the same or improved over the previous 3 years. These results are contrary to those from a recent national needs-analysis survey conducted with dairy farmers in New Zealand (Burke et al. 2008), in which respondents indicated they were generally dissatisfied with their herd's reproductive performance. This difference may be due to use of different sample-selection processes, and that in the current study, KDM were asked to rate their herd's reproductive performance before they were asked about constraints and management aspects of reproduction. KDM similarly rated the quality of genetics and milksolids performance of their herds (Figure 3).

Apparent discrepancy between evidence that the fertility of dairy cows is declining in New Zealand (Harris et al. 2006) and perceptions by KDM about reproductive performance in their herds in the current study highlights a need for advisors and extension programmes to provide objective benchmarking of herd reproductive performance (Bigraspoulin et al. 1985ab). Adult learning studies identify ‘establishing dissatisfaction with the status quo’ as a fundamental criterion for change (Holifield and Masters 1993). In the 2 years preceding the study presented here, there had been prolonged periods of drought; these had resolved in the year immediately prior to when the questionnaire was administered, and this may explain the perceived improvement in reproductive performance reported by KDM in the current study. When asked how the reproductive performance of their herd compared with that of other herds in the district, 123/133 (92%) considered their herd was either the same or better. Most based this response on conversations with other farmers or local veterinarians, and regarded their herd's final empty rate, i.e. the proportion of cows not pregnant at the end of the mating period, as the chief measure of performance. However, as the length of the mating period varies between herds, this measure is not easily comparable. That placed further responsibility on the rural professional to be able to provide appropriate benchmarking measures for comparing performance between herds, and within herds, over time, e.g. the 6‐week in-calf rate. The importance of providing a realistic frame of reference when veterinarians are advising farmers on herd health has been highlighted by Lam et al. (2011).

The perceived constraints on reproductive performance rated most likely were anoestrous cows within the herd before and during mating, protracted calving pattern, and low body condition score of the herd. Despite identifying anoestrous cows as a likely constraint to reproductive performance, most KDM considered heat detection was unlikely or very unlikely to be a constraint, similar to findings in a recent farmer needs-analysis survey (Burke et al. 2008). In that survey, respondents were offered a series of options, and prioritised balancing nutrition, managing foreign genetics, and body condition score of cows as the most important barriers to meeting reproductive targets. Within the current study, the options offered as constraints were based on barriers identified in the InCalf extension programme (Morton 2004), that provided insight into constraints perceived by farmers within that framework. Amongst the additional constraints proposed by the KDM themselves in this study, the most frequent was nutrition, consistent with the farmer needs-analysis results reported by Burke et al. (2008). The use of veterinary interventions (induction of parturition, hormonal treatment of anoestrous cows, and routine treatment for endometritis) was similar between regions, and this may have reflected similar levels of advice and service provided by veterinary practices in relation to these interventions, across regions.

When asked to prioritise areas of management for the future, ratings were consistently high, with 3/10 options having a median rating of 5 (very high priority). This could have reflected fairly homogeneous positive attitudes toward farming amongst KDM in this study. However, this may also have been due, in part, to acquiescence bias. In this study, management of reproduction was prioritised the highest of the options offered, with 89/133 (67%) respondents rating it 5/5. Of the respondents that employed staff, 87/110 (79%) rated the management of staff as a high or very high priority. However, ratings of attitudes toward the importance of staff training, choosing staff to match roles, and importance of staff suggestions were lower. A recent study in Australia highlighted that identifying the key members of the on-farm team who contributed to decisions about mastitis and milk quality, and involving them in the consultation, increased the success of a programme to improve milk quality (Penry et al. 2011). When seeking to improve herd reproductive performance, engaging all staff on the farm is likely to be a critical part of achieving improvement. The occupation of the KDM was a significant predictor of the perceived value of staff training in the multivariable models. The higher rating from managers could refect desire for training themselves as well as for their staff.

Multivariable modelling was undertaken to allow assessment of the association of various demographic variables with attitudes, priorities and constraints perceived by KDM. Overall, 30–39-and 40–49‐year-olds rated the attitudinal statements offered more highly than 20–29‐year-olds. Interestingly, the 20–29‐year-old age group also had half the odds of rating their herd reproductive performance as high or higher than any particular rating relative to all other age groups. Lower rating of performance and greater benefit perceived from group learning may make this age group more responsive to advice and extension programmes designed to improve herd reproductive performance. Also notable was that lower-order sharemilkers (KDM with ≤35% stakehold in the farm and stock) and older KDM (>29 years old) were more likely to use proven management protocols than owner-operators or younger KDM, respectively. KDM managing more intensive production systems (i.e. >2) were more likely to try unproven ideas or technologies, probably reflecting the fact that intensive dairy systems depart from the traditional dairy farm systems and practices in New Zealand.

The results from this study indicated that the age of KDM affected their attitudes, priorities and perceived constraints. Furthermore, both the 20–29- and ≥50‐year-old age groups generally reported lower ratings for many of the attitudes investigated. Such U-shaped responses could be described using age-period cohort methodology in future studies, in which the separate effects of age and cohort can be identified using a longitudinal study protocol (O'Brien 2000; Harding 2009). This methodology accounts for ‘cohort succession, ageing, and period-specific historical events to explain social and demographic change (Smith 2008). We propose that these effects of age group were both due to age and cohort, with changes in attitude differentiated by experience and maturity (Brockmann 2009), and the effect of recent economic downturn, for example, on different age groups. Extension programmes should be developed considering the age group of the KDM. Region had a lesser effect than age, but should also be considered in the development of extension programmes.

In conclusion, KDM generally rated the reproductive performance of their herds as high or very high although, paradoxically, reproduction was listed as the highest priority for future improvement. Without challenging this rating of perceived performance, uptake of reproductive advice and extension programmes, and consequent positive behavioural change, is less likely. Appropriate objective benchmarking of actual reproductive performance may help KDM understand the potential for and consequent benefits of improvement.

Extension approaches should be developed that are appropriate to age group, as it was evident that the attitudes, priorities and perceived constraints of KDM varied with age. The value of staff suggestions and training was rated lower than many other attitudes assessed. Accordingly, extension programmes that encourage KDM to focus on engaging all farm staff may be more successful. Finally, biophysical and demographic factors such as the size of the herd and the occupation of the KDM did not account for important variation in attitude. This indicates that extension approaches stratified on biophysical factors such as the size of the herd, e.g. extension programmes for a large farm vs a small farm, may not be required.

Acknowledgements

This National Herd Fertility Study was funded by dairy farmers in New Zealand through DairyNZ Inc. (AN806), Sustainable Farming Fund (Ministry of Agriculture and Forestry) and Agmardt. The KDM and regional veterinarians and technicians are gratefully acknowledged for their willing support.

Notes

¹ http://www.sciquest.org.nz/node/70039