such information may be incomplete or unable to be used in any specific situation. No ..... Since the advent of COAG's water reform agenda, and concerns associated with the ... (salt, nutrients, pesticides etc) in the waterways from runoff or drainage ..... pose, and the amount of control they would have to prevent it happening.
Towards Sustainable Irrigation Practices: Understanding the Irrigator A case study in the Riverland – South Australia Zoe Leviston, Natasha B. Porter, Bradley S. Jorgensen, Blair E. Nancarrow and Lorraine E. Bates September 2005
Copyright and Disclaimer © 2005 CSIRO To the extent permitted by law, all rights are reserved and no part of this publication covered by copyright may be reproduced or copied in any form or by any means except with the written permission of CSIRO Land and Water. Important Disclaimer: CSIRO advises that the information contained in this publication comprises general statements based on scientific research. The reader is advised and needs to be aware that such information may be incomplete or unable to be used in any specific situation. No reliance or actions must therefore be made on that information without seeking prior expert professional, scientific and technical advice. To the extent permitted by law, CSIRO (including its employees and consultants) excludes all liability to any person for any consequences, including but not limited to all losses, damages, costs, expenses and any other compensation, arising directly or indirectly from using this publication (in part or in whole) and any information or material contained in it. Cover Photograph: Description: Flooded Murray River at Mildura, VIC Photographer: Willem van Aken © 2005 CSIRO
Towards Sustainable Irrigation Practices: Understanding the Irrigator A case study in the Riverland – South Australia Zoe Leviston, Natasha B. Porter, Bradley S. Jorgensen, Blair E. Nancarrow and Lorraine E. Bates Australian Research Centre for Water in Society
CSIRO Land and Water Client Report September 2005
Acknowledgements This work was commissioned and funded by the Department of Water, Land and Biodiversity Conservation (DWLBC), and the River Murray Catchment Water Management Board, South Australia. The authors gratefully acknowledge the ongoing advice and assistance of the following people: John Rolls, Lisa Stribley and Ingrid Franssen - Department of Water, Land and Biodiversity Conservation Dan Meldrum - River Murray Catchment Water Management Board Jeff Parish, Gavin McMahon and Jim Atsaves - Central Irrigation Trust David Morris and Mirco De Col - Renmark Irrigation Trust
Table of Contents 1.0
INTRODUCTION ..................................................................................... 1
2.0
RESEARCH METHODOLOGY ............................................................... 2
2.1 AJZEN’S THEORY OF PLANNED BEHAVIOUR ....................................................... 2 2.2 SCOPING INTERVIEWS ....................................................................................... 3 2.2.1 Methodology ............................................................................................................ 3 2.3 THE HYPOTHESISED MODEL.............................................................................. 4 2.3.1 Behaviour................................................................................................................. 5 2.3.2 Attitudes.................................................................................................................. 5 2.3.3 Subjective Norm ...................................................................................................... 5 2.3.4 Perceived Control .................................................................................................... 5 2.3.5 Risk.......................................................................................................................... 6 2.3.6 Trust........................................................................................................................ 6 2.3.7 Responsibility .......................................................................................................... 6 2.3.8 Values ...................................................................................................................... 6 2.4 TESTING THE MODEL ........................................................................................ 7 2.4.1 Study Area and Participants................................................................................... 7 2.4.2 The Questionnaire ................................................................................................... 8
3.0
RESULTS................................................................................................ 8
3.1 PRELIMINARY ANALYSIS .................................................................................... 9 3.1.1 Pressure to use less water........................................................................................ 9 3.1.2 Influencing water use for irrigation ...................................................................... 10 3.1.3 Using less water .................................................................................................... 12 3.1.4 Risk........................................................................................................................ 14 3.1.5 Trust..................................................................................................................... 18 3.1.6 Attitudinal Statements......................................................................................... 21 3.1.7 Demographics........................................................................................................ 24 3.2 THE STRUCTURAL EQUATION MODEL ............................................................. 31 3.2.1 Constructing the Behaviour Measure ................................................................... 31 3.2.2 Running the Model ............................................................................................... 37 3.2.3 Socio‐Demographic Analyses ................................................................................ 45
4.0
DISCUSSION AND RECOMMENDATIONS ......................................... 46
5.0
REFERENCES ...................................................................................... 50
APPENDIX 1 .................................................................................................... 53 APPENDIX 2 .................................................................................................... 57 APPENDIX 3 .................................................................................................... 61 APPENDIX 4 .................................................................................................... 73 APPENDIX 5 .................................................................................................... 77
1.0
INTRODUCTION
Since the advent of COAG’s water reform agenda, and concerns associated with the possible effects of climate change, the use of Australia’s water resources has come under increasing scrutiny. This is particularly the case in rural areas, where irrigators are now being questioned over the proportion of the nation’s water being consumed by that industry. Catchment managers are faced with the environmental effects of increasing contaminants (salt, nutrients, pesticides etc) in the waterways from runoff or drainage associated with excessive irrigation water. Both rivers and groundwater aquifers are suffering environmental degradation through lowering water levels. Irrigators themselves are feeling increasingly under scrutiny and unappreciated by urban Australia. In the past, much has been tried to encourage the uptake of innovative technology and practices from extension and education, through participatory approaches and management plans to economic incentives and disincentives. These have generally not resulted in the widespread changes that were hoped for. However, what is not known is how irrigators make their decisions to take up technologies and incentives, and whether these initiatives are meeting the needs of irrigators. Issues such as the role of local knowledge, lifestyle, trust in authorities, and risk in irrigators’ decision‐making are generally not considered in programs aimed at promoting water efficient practices and technologies. Until it is understood what values, beliefs and attitudes underpin irrigators’ decisions to accept or reject water efficient practices, it is unlikely that the desired sustainability targets will be achieved. It is essential for governments, industry and science to better understand the irrigators and their particular needs so they can better design technologies and communication programs that will provide mutual benefit. This study aimed to identify the key individual psychological and social factors in irrigators’ decisions to adopt or reject improved irrigation practices in the South Australian Riverland. Once these factors are established, it will be possible to identify and recommend available measures to better align irrigation efficiency developments, investigations and communications with the circumstances and needs of the irrigator, to encourage greater uptake of sustainable practices. This research was conducted by the CSIRO Australian Research Centre for Water in Society (ARCWIS) for the Department of Water, Land and Biodiversity Conservation (DWLBC) South Australia, and the River Murray Catchment Water Management Board.
1
2.0
RESEARCH METHODOLOGY
To date, most studies dealing with the irrigation behaviour of farmers in Australia have dealt with the impacts of particular programs, policies and initiatives by investigating broad‐scale social and community factors that may influence generalised ‘pro‐environmental’ behaviour (eg. Crean, Shaw, Singh & Mullen, 2004; Burrows & Boland, 2002; Prior, 2003; Grasby, Lockie & McAllister, 2000; Kraak, 2000). This study aims to investigate the specific psychological factors that predict the adoption of a specific behaviour. A small number of studies have tried to pinpoint predictive factors for the adoption of sustainable practices by Australian farmers, most notably Crase and Maybery (2002). They hypothesised that farmers’ personalities, values, attitudes to conservation and a range of demographic variables would be important contributors to land stewardship decisions. While they found that certain attitudinal and lifestyle factors played a part, personality had little to no influence over adoption, and where influence was exerted it was acknowledged that personality was beyond the realm of resource management programs to alter. Notably absent from these studies has been the components of perceived control and subjective norms, which a large and growing body of research suggests are critical in predicting the adoption of ‘pro‐environmental’ behaviours (eg. Terry, Hogg & White, 1999; Harland, Staats & Wilke, 1999; Kalafatis, Pollard, East & Tsogas, 1999; Bamberg 2002; Lam, 1999).
2.1
Ajzen’s Theory of Planned Behaviour
Ajzen’s Theory of Planned Behaviour (1985), see Figure 1 below, proposes that a person’s actual behaviour can be predicted from their behavioural intention. This intention is in turn determined by a person’s attitudes towards performing that particular behaviour; subjective norm (whether or not most people important to the person think the behaviour should be complied with); and perceived behavioural control (the perceived ease or difficulty of performing the action).
Behavioural beliefs Outcome evaluations
Attitudes
Normative beliefs Motivation to comply
Subjective Norms
Control beliefs Personal power
Perceived Control
Intended Behaviour
BEHAVIOUR
Figure 1: Ajzen’s Theory of Planned Behaviour
2
This theory has been used effectively to predict behaviour in a wide variety of contexts (eg. physical activities, quitting cigarette smoking, blood donating, internet use and so on). Ajzen’s theory has also been used overseas to predict the irrigation behaviour of farmers (eg. Lynne, Casey, Hodges & Rahamni, 1995). The Theory of Planned Behaviour has been chosen as a base model in this study as it is a well‐tested and widely acknowledged robust basic framework for building more context‐ relevant models relating to the determinants of performing a specific behaviour (East, 1997; Manstead & Parker, 1995). Also, ARCWIS has used the Theory of Planned Behaviour as a theoretical basis to successfully model the prediction of community behaviour in relation to wastewater reuse (Po, Nancarrow, Leviston, Porter, Syme & Kaercher, 2005).
2.2
Scoping Interviews
A series of semi‐structured interviews were conducted by the study team with a range of local irrigators and key stakeholders in the Riverland to gain an understanding of the local context. The interviews, undertaken in November 2004, were designed to elicit irrigator attitudes, beliefs and values in relation to water use, as well as identifying other possible psychologically‐ or socially‐based variables that may influence their decisions and behaviours. The findings of these interviews were used to better define the variables in Ajzen’s model and identify additional variables that may influence decision‐making and behaviour. 2.2.1 Methodology Names of key stakeholders and a representative range of irrigators were compiled through discussions with local contacts and through recommendations received over the course of the interviews. The aim was to cover a broad range of perspectives. Interviews were arranged in advance and a confirmation letter sent to briefly explain the purpose of the visit (see Appendix 1 for an example letter). Two teams of two study personnel visited interviewees in their preferred locations (frequently on their properties) and conducted interviews in a relaxed and informal manner. Many were single interviews, but a number were in pairs or small groups of no more than five people. The semi‐structured interviews followed a checklist as shown in Appendix 2. Conversations were allowed to flow freely, while ensuring that all topics had been covered by the conclusion of the interview. Discussions ranged from forty‐five minutes to two hours in duration. A total of thirty‐nine people were interviewed in the Riverland during thirty separate interviews. A brief summary of these interviews is included as Appendix 3. As the interviews progressed, a number of themes emerged that were to inform the structure of the hypothesised model.
3
2.3
The Hypothesised Model
After consideration of the findings from the scoping interviews, an assessment of similar studies (both national and international), and a review of findings from previous water‐use research carried out by ARCWIS, the variables for the hypothesised model were determined. As issues of perceived control and personal power were cited frequently (and unprompted) throughout the scoping interviews, Ajzen’s model remained to form the theoretical basis of this work. In addition, a number of further variables were incorporated. Risk and responsibility were two factors that emerged strongly from the scoping interviews, as did lifestyle and value issues. Previous research supports the notion that lifestyle and values are important predictors of pro‐environmental behaviour among Australian irrigators (Crase & Maybery, 2002; Kraak, 2000). Trust in irrigation information from a variety of sources (notably government agencies and private consultants) was a topic touched on by a number of those interviewed. The importance of trust has been established in previous research undertaken by ARCWIS with regards to urban water services, and is an emerging issue rurally (Porter, Nancarrow & Syme, 2004; Porter, Leviston, Nancarrow, Po & Syme, 2005; Po et al., 2005). Figure 2 below outlines the hypothesised model which formed the basis of the irrigator questionnaire. Outcome Evaluations
Attitudes
Behavioural Beliefs
Subjective Norm
Normative Beliefs Motivation to comply
Perceived Control
Control Beliefs Personal Power
Risk Perception
Risk
Risk Behaviour
Trust in authorities, technology and information
Behaviour
Trust
Individual, community and authorities’ responsibility for water security
Responsibility
Lifestyle Anthropocentric Beliefs
Values © 2004 CSIRO – Commercial -in-confidence
Figure 2: Hypothesised Model – Understanding the Irrigator
4
The following sections provide an overview of the hypothesised model’s components. 2.3.1 Behaviour The Behaviour component refers to the individual irrigator’s water using behaviour. It is measured by comparing recommended levels of water use for the individual’s crop with their actual water consumption.1 This is referred to as the “water use index”2 throughout the report. The target behaviour referred to in the model components below is ‘using less water for irrigation purposes’. 2.3.2 Attitudes The Attitudes component refers to the favourableness of the irrigator’s evaluations of the behaviour. It is measured by: • behavioural beliefs: what the irrigator thinks the outcomes of adopting the behaviour will be; and • outcome evaluations: how favourable these outcomes are to the irrigator. 2.3.3 Subjective Norm The Subjective Norm component refers to the extent to which people and groups important to the irrigator think the behaviour should be adopted. It is measured by: • normative beliefs: the irrigator’s perception of whether others think the behaviour should be adopted; and • motivation to comply: how influential these others are in decision‐making processes surrounding the adoption of the behaviour. 2.3.4 Perceived Control The Perceived Control component refers to the perception of how easy or difficult it is for the irrigator to adopt the behaviour. It is measured by: • control beliefs: the perceived control over the decision to adopt the behaviour; and • personal power: the perceived power to overcome factors that may hinder the adoption of the behaviour. 1
For this study, as we were able to obtain actual data for the target behaviour, we were able to dispense with the ‘Intended Behaviour’ variable from Ajzen’s original model. Thus, links between predictor variables and the target behaviour can be made directly. 2 While it is admitted that there are problems associated with the calculation of recommended water use for each crop (eg. soil types, etc), the water use index was considered to be the best way of allowing comparison and analysis across crop types. 5
2.3.5 Risk The Risk component refers to the perceived risk associated with adopting the behaviour. It is measured by: • risk perception: the perceived likelihood and seriousness of negative consequences associated with the adoption of the behaviour; and • risk behaviour: the irrigator’s propensity to view risks as acceptable 2.3.6 Trust The Trust component refers to the amount of trust the irrigator places in authorities, information and technology to promote water use efficiency. It is measured by assessing trust in relation to information regarding the behaviour provided by specific organisations and groups. 2.3.7 Responsibility The Responsibility component refers to the irrigator’s belief in the relative responsibilities of the individual, community and authorities in relation to the behaviour. It is measured by assessing the responsibility of the irrigator, the community and government for water security in the region. 2.3.8 Values The Values component refers to the irrigator’s personal values that influence the behaviour. It is measured by: • lifestyle: the place farming occupies in the irrigator’s lifestyle preferences; and • anthropocentric beliefs: the irrigator’s beliefs in the existence of water primarily for human use. 6
2.4
Testing the Model
The hypothesised model was tested using data collected via a telephone survey with irrigators in the Riverland region in May‐June 2005. A sample size of greater than 450 was required for the preferred data analysis technique, structural equation modelling. 2.4.1 Study Area and Participants Irrigators from across the Riverland region were selected to be involved in the survey. Contact details of irrigators in the Riverland were provided by the DLWBC (Private Diverters), the Central Irrigation Trust (CIT) and the Renmark Irrigation Trust (RIT). Contacts were sought for the Sunlands/Qualco Irrigation Area via the White and Yellow Pages. Original intentions were to sample equally from the Private Diverter Irrigators, and Renmark and Sunlands/Qualco Irrigation areas to ensure a good spread across the area. On advice from the CIT, RIT and Sunlands/Qualco representatives, and the River Murray Catchment Water Management Board, a larger proportion of irrigators from the Central Irrigation area were sought. Where information on property‐size was available (RIT and Private Diverters), a cross‐section of small, medium and large growers were targeted to ensure the sample included a range of viewpoints. In total 509 irrigators were surveyed. Due to a number of implementation issues the final number of respondents from each of the areas was as follows. Table 1. Number of respondents surveyed based on source of water n (509)
%
Central Irrigation Trust
263
52.1
Private Diverter Irrigators
96
18.9
Renmark Irrigation Trust
97
19.1
Sunlands/Qualco Irrigation Trust
51
10.0
A trained team of interviewers administered the survey via telephone and were directed to obtain the person who made decisions regarding irrigation on the target property for each interview. Interviewers were further instructed to contact each property on their lists at least five times, at different times of the day, before the property could be classed as a ‘non‐contact’. The refusal rate for the survey was 35.9%.3 The following table gives a breakdown of refusal reasons.
3
This figure is quite low given current refusal trends for telephone surveys. 7
Table 2. Refusal details n (273)
Reason Not Interested
108
Too Busy
81
Limited English
62
Hung Up
11
Unwell
9
Elderly
2
2.4.2 The Questionnaire Based on the hypothesised model, the literature and the information and language provided in the scoping interviews, a draft questionnaire was developed and pre‐tested. As a result of pre‐testing, a number of minor changes were made to the questionnaire before being finalised.
3.0
RESULTS
Preliminary analyses were undertaken using correlation, analysis of variance (ANOVA), factor analysis and reliability analysis. This was followed by investigation of the causal relationships between the components of the model using the robust maximum likelihood estimation method in LISREL 8.72 (Joreskorg, Sorbom, du Toit & du Toit, 2000). For the preliminary analyses, a significance level of p 50 years
37
7.3
Allocation group
Mean = 24.25 years
25
Reported current water allocation Respondents were asked what their current water allocation was. Responses are summarised in the table below. Table 23. Respondents’ current water allocation (ML) n (509)
%
4 – 50
110
21.6
51 – 150
178
35.0
151 – 350
129
25.3
351 – 700
40
7.9
701 – 2000
19
3.7
2001 - 7000
7
1.4
Unknown
26
5.1
Allocation (ML)
Mean = 268.15
Reported average annual water use Respondents were asked how much water they used on average per year for irrigation. Their responses are summarised in the table below. Table 24. Respondents’ reported annual average water use (ML) n (509)
%
1 – 50
153
30.1
51 – 150
181
35.6
151 – 350
104
20.4
351 – 700
24
4.7
701 – 2000
10
2.0
2001 - 7000
7
1.4
Unknown
30
5.9
Annual Water Use (ML)
Mean = 210.29
26
Size of crop under irrigation Respondents were asked the size of their various crops under irrigation. Their responses are summarised in the table below. Table 25. Respondents’ reported size of crops (Ha) n (506)
%
0.4 – 5
95
18.8
5 – 10
115
22.7
10 – 20
127
25.1
20 – 50
120
23.7
50 – 150
34
6.7
150 – 500
13
2.6
500 – 1280
2
0.4
Hectares
Mean = 28.87
Irrigation methods Respondents were asked what irrigation methods they used on each of their crops. Many respondents (41.9%) had multiple irrigation methods. The most common, as reported by respondents, are listed below. Table 26. Most common irrigation methods used on crops as reported by respondents n (509)
%
Drip
160
31.4
Overhead Sprinklers
149
29.3
Under tree Sprinklers
131
25.7
Under vine Sprinklers
85
16.7
Low throw Sprinklers
83
16.3
Waterbird Sprinklers
56
11.0
Micro – Jet
46
9.0
Irrigation Method
A full list of irrigation methods can be seen in Appendix 5.
27
Source of income Respondents were asked whether farming was their main source of income. The majority (78.8%) of respondents replied yes, compared to 21.4% who replied no. For those whose main source of income was not farming, the most common sources of income are listed in the table below. Table 27. Most common sources of income other than farming n (113)
%
Irrigation Sales and Consultancy
9
8.0
Truck Driver
6
5.3
Contracting
7
6.2
Fruit Packer/Picker
6
5.3
Teacher
5
4.4
Winery
5
4.4
Income
Role on Property Respondents were asked which of three options best described their role on the property, owner, manager or other. Table 28. Role respondents play on the property Role*
n (509)
%
Owner
484
95.1
Manager
24
4.7
Other – Foreman
1
0.2
* If the respondent was both Owner and Manager, Owner took precedence.
Respondents were also asked how long they planned to continue in this role.
28
Table 29. Length of time respondents planned to continue in current role n (505)
%
Get out as soon as possible
14
2.8
1 – 5 years
59
11.7
6 – 10 years
49
9.7
11 – 20 years
46
9.1
21 – 30 years
10
2.0
More than 30 years
6
1.2
Forever
263
52.1
Don’t know
58
11.5
Number of years
Country of birth Respondents were asked which country they were born in. Responses are listed below. Table 30. Respondents’ Country of Birth Country
n (503)
%
Australia
440
87.5
Greece
12
2.4
England
12
2.4
India
10
2.0
Germany
7
1.4
Croatia
3
0.6
Cypress
3
0.6
Italy
3
0.6
Netherlands
3
0.6
New Zealand
2
0.4
Spain
2
0.4
Turkey
2
0.4
Hungary
1
0.2
South Africa
1
0.2
USA
1
0.2
Vietnam
1
0.2
29
Age The following table provides a breakdown of respondents’ age groups. Table 31. Number of respondents in each age group n (509)
%
18 to 24 years
3
0.6
25 to 39 years
96
18.9
40 to 55 years
256
50.2
56 to 65 years
123
24.2
66 to 75 years
26
5.1
More than 75 years
5
1.0
Age Group
Education Table 32 provides a breakdown of respondents’ highest completed levels of formal education. Table 32. Details of respondents’ highest levels of formal education n (509)
%
All or some of primary school
28
5.5
All or some of secondary school
304
59.7
Trade or technical qualification
91
17.9
Agricultural Qualification
42
8.3
University qualification
44
8.6
Education
Respondents with an Agricultural Qualification were asked for details regarding their qualification. Of the forty‐two respondents who said they had Agricultural Qualifications, thirty‐three provided details. 30
Table 33. Details of respondents’ agricultural qualification Agricultural Qualification
n (42)
Diploma of Horticulture
9
Short Agricultural Courses
5
Horticultural Management Practices
4
Advanced Diploma of Horticulture
3
Diploma of Agriculture
3
Vine and Citrus Diploma
2
Rural Business
1
Bachelor of Agriculture
1
Associate Diploma in Wine-making
1
Viticulture/crop Management
1
Water Management
1
Diploma Parks and Wildlife Management
1
Farm Management
1
3.2
The Structural Equation Model
Structural equation modelling with latent variables was used to develop a model using as its theoretical basis the hypothesised model described in Section 2.3. Structural equation modelling allows for both latent and observed variables to be represented. This allowance for the simultaneous and holistic analysis of the entire system of variables makes structural equation modelling the preferred method of analysis at this stage of the research program. 3.2.1 Constructing the Behaviour Measure To construct a behaviour measure, respondents were asked to provide their average annual Mega litre (ML) usage in association with the size of their crops under irrigation. Due to anticipated inaccuracies with self‐reporting, actual usage was sought for as many respondents as possible. Actual usage was received for most respondents in the CIT and RIT areas, as well as for a majority of private diverter respondents.4 In all, actual usage was obtained for 73.7% of respondents (n=375). To check the accuracy of self‐reporting, a ‘discrepancy percentage’ was established by calculating the percentage of difference between self‐reported usage figures and actual usage figures received from the trusts (for the cases where both figures were available). The median discrepancy percentage was 6.7%, indicating that more people overestimated their 4
Respondents provided written permission to obtain this usage data for 2003/04. 31
water use than underestimated. Forty‐two percent of respondents estimated their water‐ usage to within 10% of the figure provided by the trusts. These respondents were identified as “accurate reporters” for the purposes of analyses, while the remaining 58% were classified as “inaccurate reporters”.5 Of the inaccurate reporters, 74.6% overestimated their water usage, while only 25.3% underestimated their water usage. Chi‐square tests and independent samples t‐tests were performed to identify any significant relationships between accuracy of reporting and the following variables: • Crop type • Size of crop under irrigation • Years the respondent had had their water allocations • Whether the respondents thought that they could use less water than currently • Perceived problems with trying to use less water • Perceived possibility of something going wrong as a result of using less water • Whether the respondents thought that the problems of using less water outweighed the benefits, or vice versa • ‘Water Use Index’ (see below) No significant relationships were found, suggesting that any discrepancy between reported and actual usage was not associated with any particular demographic group or particular view. Further comparison between respondents’ reported usage and their actual usage revealed a correlation of .98. Due to the strength of this correlation, it was decided that reported usage could be used in place of real usage, thereby increasing the sample size and with it, the robustness of any model produced.6 Where the respondents did not know their usage, actual usage figures were substituted. There were 20 cases where neither reported usage nor actual usage figures were available. These cases were omitted in the modelling analyses, leaving a sample size of 489. To arrive at a behaviour measure that would take into account the differing water needs of different crops, an “optimal” usage figure was calculated by applying crop water usage factors provided by the Department of Land, Water and Biodiversity Conservation to each crop of the respondents. In this manner, an overall “optimal” amount of water was calculated for each respondent based on crop type/s and the area under irrigation. The respondents’ reported usage was then divided by their “optimal” usage to arrive at a ‘water use index’. Here, a score of more than 1 signified that the respondent was using “more” water, and a score of less than 1 signified that the respondent was using “less” water.7 5
A number of these discrepancies may be accounted for by different time-frame references. While annual usage figures received from the trusts referred specifically to the ‘03/’04 financial year, respondents were asked to report their “average annual usage”. Additionally, it is plausible that in some cases the grower was referring to multiple properties, while usage figures received from the trust may have referred to only one or some of these properties. 6 This strong correlation tells us that the ranking order of nearly all the cases would be the same whether using ‘actual’ or ‘reported’ usage as the dependent variable, and hence would make no difference to the structural equation modelling. 7 It is recognized that the “optimal” crop water use figures are highly controversial with many considering that they do not account for the range of variables involved in crop water requirements. However, they provide the only means to be able to analyse water use behaviours of growers of different crops and crop combinations. For 32
Reported usage results and water use index results for all respondents can be seen in Figures 3 and 4. 160
140
Number of respondents
120
100
80
60
40
20
>3 0
28 -30
26 -28
24 -26
22 -24
20 -22
18 -20
16 -18
14 -16
12 -14
10 -12
8-1 0
6-8
4-6
2-4
0-2
0
ML per hectare
Figure 3: Water use per hectare for all respondents across crop types
80
70
Number of respondents
60
50
40
30
20
10
4.1
4
3.2
2.1
2
1.9
1.8
1.7
1.6
1.5
1.4
1.3
1.2
1.1
1
0.8
0.9
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Water use index
Figure 4: Water use index for all respondents across crop types this reason, the calculation is referred to as a ‘water use index’ and provides a range of water use (from lesser to greater) for consideration in the analysis. No judgment is made about what indicates favourable water using behaviour. 33
In addition to whole group analyses, separate analyses were performed on each of the major crop types. Here, only those cases where one type of crop was grown was selected for inclusion (thereby excluding those respondents with multiple crop types) as water use data was only available for the whole property and not for individual crop areas. Figures 5 and 6 give water usage and water use index results for those who grew grapes exclusively8 (n=196). 80
70
Number of respondents
60
50
40
30
20
10
20 -2 2
80 -2 0
16 -1 8
14 -1 6
12 -1 4
10 -1 2
810
68
46
24
02
0
ML per Hectare
Figure 5: Water use per hectare for grape growers
40
35
Number of respondents
30
25
20
15
10
5
2. 2
2. 1
2. 0
1. 9
1. 8
1. 7
1. 6
1. 5
1. 4
1. 3
1. 2
1. 1
1. 0
0. 9
0. 8
0. 7
0. 6
0. 5
0. 4
0. 3
0. 2
0. 1
0
W ater use index
Figure 6: Water use index for grape growers 8
No discerning data was available for different water requirements for grapes for bulk wine, grapes for quality wines, table grapes and so on. 34
Figures 7 and 8 give water usage and water use index results for those who grew citrus exclusively (n=44).
16
14
Number of respondents
12
10
8
6
4
2
6 >2
24
-2
6
4 -2 22
20
-2
2
0 -2 18
-1 16
14
12
8
6 -1
4 -1
2
8 6-
-1 10
6 4-
10
4 2-
8-
2 0-
0
ML per Hectare
Figure 7: Water use per hectare for citrus growers
9
8
7
Number of respondents
6
5
4
3
2
1
W ater use index
Figure 8: Water use index for citrus growers
35
2. 1
1. 7
1. 3
1. 2
1. 1
1. 0
0. 9
0. 8
0. 7
0. 6
0. 5
0. 4
0. 3
0. 2
0. 1
0
Figures 9 and 10 give water usage and water use index results for those who grew stonefruit exclusively (n=21). 6
5
Number of respondents
4
3
2
1
8
6
-1 16
12
14
-1
-1
4
2
8 6-
-1 10
6 4-
10
4 2-
8-
2 0-
0
ML per Hectare
Figure 9: Water use per hectare for stonefruit growers
4.5
4
3.5
Number of respondents
3
2.5
2
1.5
1
0.5
1. 2
1. 1
1. 0
0. 9
0. 8
0. 7
0. 6
0. 5
0. 4
0. 3
0. 2
0. 1
0
W ater use index
Figure 10: Water use index for stonefruit growers
36
Outliers The final data set contained a number of outliers for the water use and water use index variables. That is, a small handful of scores appeared to deviate markedly from the bulk of the scores in these variables. In these cases, individual questionnaires were checked to ensure no error had occurred during the coding or data entry phases. As a number of outliers remained after this process, water usage and water use index figures were mathematically modified by applying a logarithm to normalise their distributions. In this way, genuine outliers were able to be retained without violating the assumptions of normality that are inherent in some of the statistical analyses that were applied. 3.2.2 Running the Model The structural equation model in Figure 11 was estimated using LISREL 8.72 software and Robust Maximum Likelihood estimation (Joreskog et al, 2000). The model can be interpreted by referring to its two main parts. The first part of the model refers to the relationships between the latent variables (shown in the model as ellipses) and their respective indicators 9 (shown in the model as rectangles). Put simply, this aspect of the model answers the question ‘How well do the indicators measure the latent variables of interest?’ Coefficients on these paths can range from ‐1.0 (i.e., a strong negative relationship between the latent variable and the indicator) and +1.0 (i.e., a strong positive relationship between the latent variable and the indicator. Figure 11 shows that all indicators in the model have strong positive relationships with the latent variables they were hypothesised to measure. In fact, all of these relationships are statistically significant at the .001 level. Note, however, that it was necessary in the case of latent variables measured with only one indicator to fix these paths to a pre‐specified value. In most cases, single indicators were assumed to correlate 0.84 with their respective latent variables. The two exceptions were the indicators for Age (AGE1) and whether Farming was the respondent’s main source of income (FARMINC). These indicators were assumed to have been reliably measured given that they were unlikely to have been difficult to answer for respondents. The second part of the structural equation model refers to the relationships between the independent latent variables and the dependent variable (i.e., Water‐use Behaviour, measured here by the water use index). The coefficients on these paths are free to range from ‐1.0 (i.e., a strong negative relationship between the predictor and Water‐use Behaviour) and +1.0 (i.e., a strong positive relationship between the predictor and Water‐use Behaviour). Note, however, that the path between Trust in Water Efficiencies and Water‐use Behaviour was not estimated in the model as it was highly correlated (.89, p.05
SRMR
.033
≤ .08
TLI
.99
≥ .90
GFI
.93
≥ .90
.016 (90%CI = .007, .022)
≤ .08
Chi-square (df)
RMSEA
40
Not shown in Figure 11 are the zero‐order correlations among the predictors and Water‐use Behaviour. These correlations revealed that few variables in the model had significant relationships with Water‐use Behaviour. The largest correlations were for Outcome Evaluation of Using Less Water (‐.19, p