Role of input self-sufficiency in the economic and environmental sustainability of specialised dairy farms

Abstract

Increasing input self-sufficiency is often viewed as a target to improve sustainability of dairy farms. However, few studies have specifically analysed input self-sufficiency, by including several technical inputs and without only focussing on animal feeding, in order to explore its impact on farm sustainability. To address this gap, our work has three objectives as follows: (1) identifying the structural characteristics required by specialised dairy farms located in the grassland area to be self-sufficient; (2) analysing the relationships between input self-sufficiency, environmental and economic sustainability; and (3) studying how the farms react to a decrease in milk price according to their self-sufficiency degree. Based on farm accounting databases, we categorised 335 Walloon specialised conventional dairy farms into four classes according to their level of input self-sufficiency. To this end, we used as proxy the indicator of economic autonomy - that is, the ratio between costs of inputs related to animal production, crop production and energy use and the total gross product. Classes were then compared using multiple comparison tests and canonical discriminant analysis. A total of 30 organic farms - among which 63% had a high level of economic autonomy - were considered separately and compared with the most autonomous class. We showed that a high degree of economic autonomy is associated, in conventional farms, with a high proportion of permanent grassland in the agricultural area. The most autonomous farms used less input - especially animal feeding - for a same output level, and therefore combined good environmental and economic performances. Our results also underlined that, in a situation of decrease in milk price, the least autonomous farms had more latitude to decrease their input-related costs without decreasing milk production. Their incomes per work unit were, therefore, less impacted by falling prices, but remained lower than those of more autonomous farms. In such a situation, organic farms kept stable incomes, because of a slighter decrease in organic milk price. Our results pave the way to study the role of increasing input self-sufficiency in the transition of dairy farming systems towards sustainability. Further research is required to study a wide range of systems and agro-ecological contexts, as well as to consider the evolution of farm sustainability in the long term.

Full text

Role of input self-sufficiency in the economic and environmental
sustainability of specialised dairy farms
T. Lebacq
1,2†
, P. V. Baret
1
and D. Stilmant
2
1
Earth and Life Institute, Université catholique de Louvain, Croix du Sud 2, L7.05.14, 1348 Louvain-la-Neuve, Belgium;
2
Centre wallon de Recherches agronomiques,
Unité Systèmes agraires, Territoire et Technologies de l’information, Rue de Serpont 100, 6800 Libramont, Belgium
(Received 14 January 2014; Accepted 23 September 2014; First published online 28 November 2014)
Increasing input self-sufficiency is often viewed as a target to improve sustainability of dairy farms. However, few studies have
specifically analysed input self-sufficiency, by including several technical inputs and without only focussing on animal feeding, in
order to explore its impact on farm sustainability. To address this gap, our work has three objectives as follows: (1) identifying
the structural characteristics required by specialised dairy farms located in the grassland area to be self-sufficient; (2) analysing the
relationships between input self-sufficiency, environmental and economic sustainability; and (3) studying how the farms react to a
decrease in milk price according to their self-sufficiency degree. Based on farm accounting databases, we categorised 335 Walloon
specialised conventional dairy farms into four classes according to their level of input self-sufficiency. To this end, we used as proxy
the indicator of economic autonomy –that is, the ratio between costs of inputs related to animal production, crop production and
energy use and the total gross product. Classes were then compared using multiple comparison tests and canonical discriminant
analysis. A total of 30 organic farms –among which 63% had a high level of economic autonomy –were considered separately
and compared with the most autonomous class. We showed that a high degree of economic autonomy is associated, in
conventional farms, with a high proportion of permanent grassland in the agricultural area. The most autonomous farms used less
input –especially animal feeding –for a same output level, and therefore combined good environmental and economic
performances. Our results also underlined that, in a situation of decrease in milk price, the least autonomous farms had more
latitude to decrease their input-related costs without decreasing milk production. Their incomes per work unit were, therefore, less
impacted by falling prices, but remained lower than those of more autonomous farms. In such a situation, organic farms kept
stable incomes, because of a slighter decrease in organic milk price. Our results pave the way to study the role of increasing input
self-sufficiency in the transition of dairy farming systems towards sustainability. Further research is required to study a wide range
of systems and agro-ecological contexts, as well as to consider the evolution of farm sustainability in the long term.
Keywords: input self-sufficiency, dairy farming, economic sustainability, environmental sustainability
Implications
Dairy farming systems are facing major changes and uncer-
tainty related to price volatility, socio-cultural values and
political aspects. At the same time, they are considered as
exerting pressure on the environment and animal welfare.
Consequently, there is a social demand for developing
alternative farming systems. Increasing input self-sufficiency
constitutes a possible pathway to design systems that are
more sustainable and able to operate in this changing
context. In this perspective, it is crucial to understand the role
of input self-sufficiency in the sustainability of dairy farms,
using an indicator based on several technical inputs –for
example, animal feeding, fertilisers and energy.
Introduction
In recent decades, the development of intensive and specialised
livestock farming systems has been called into question due to
detrimental effects on animal welfare and the environment
(Ten Napel
et al.
, 2011). Moreover, farmers now have to
operate in a context characterised by unprecedented change
and high uncertainty, such as volatility in agricultural product
prices, increases in production costs, changes in agricultural
policies and socio-cultural values (Astigarraga and Ingrand,
2011; Dumont
et al.
, 2013). As a consequence, there is
currently lively scientific and public debate about the future
evolution of livestock production (Bernués
et al.
, 2011) and the
development of alternative systems (Dumont
et al.
, 2013).
Several authors agree that increasing self-sufficiency
provides one way to develop more sustainable agricultural
†
E-mail: theresa.lebacq@uclouvain.be
Animal
(2015), 9:3, pp 544–552 © The Animal Consortium 2014
doi:10.1017/S1751731114002845
animal
544

systems in such an uncertain context (López-Ridaura
et al.
,
2002; Vilain, 2008; Bernués
et al.
, 2011). Broadly defined,
self-sufficiency is ‘
the capacity of the system to regulate
and control its interaction with the environment
’(Bernués
et al.
, 2011). Farm self-sufficiency can be considered at
three levels as follows (Ruiz
et al.
, 2011): (1) decision-
making; (2) financial –that is, related to subsidies and debts;
and (3) technical –that is, related to the use of external inputs.
Our paper specifically focusses on this last level –input self-
sufficiency –that Vilain (2008) defined as ‘
the capacity of a farm
to produce goods and services from its own resources, i.e., with
a minimal amount of external inputs
’.
Input self-sufficiency is known to have economic, environ-
mental and societal assets. First, in the coming years, the
agricultural sector will probably be confronted with an
increase in input and energy prices, because of competition
for various land uses (
feed, food, fuel
) and depletion of
oil resources (Bernués
et al.
, 2011). In this context, the
most self-sufficient systems will keep lower production costs,
giving them a comparative economic advantage. Indeed,
systems that are less dependent on inputs are less affected
by resource scarcity and price volatility (Bernués
et al.
, 2011).
Second, because of a lower consumption of inputs such as
mineral fertilisers, pesticides and animal feed, self-sufficient
systems also have a lower impact on the environment
(Vilain, 2008; Raveau, 2011). Finally, from a societal point of
view, input self-sufficiency, especially regarding animal
feeding, improves the traceability of the products (Paccard
et al.
, 2003).
In the literature, input self-sufficiency has often been used
as an attribute of sustainability in livestock farming system
analyses (e.g. López-Ridaura
et al.
, 2002; Ripoll-Bosch
et al.
,
2012). It has also been considered as a key principle of agro-
ecology for animal systems (Dumont
et al.
, 2013), as well as
a strategy to improve their resilience (Darnhofer, 2010).
However, few studies have focussed on input self-sufficiency
of dairy farms, although dairy production usually depends on
many inputs (Thomassen
et al.
, 2009). Moreover, existing
studies have analysed feed self-sufficiency without including
other inputs such as mineral fertilisers or veterinary products
(see, for instance, Paccard
et al.
, 2003). Therefore, some
issues related to input self-sufficiency of dairy farming sys-
tems need to be assessed in an objective and reproducible
way –for example, what are the relationships between input
self-sufficiency and economic, environmental and social farm
performances? Does input self-sufficiency involve specific
structural characteristics? How do self-sufficient farms react
to external changes –for example, changes in milk price?
To address these issues, this work has the following three
objectives: (1) to identify structural characteristics required
by specialised dairy farms located in the grassland area to be
self-sufficient; (2) to analyse the relationships between input
self-sufficiency, economic and environmental sustainability,
as well as to explore whether self-sufficiency allows the
farms to conciliate these two sustainability dimensions; and
(3) to study the economic impact of a decrease in milk price
according to input self-sufficiency. In this article, we address
these three objectives through the analysis of the input
self-sufficiency of 365 specialised dairy farms located in the
grassland area of Wallonia, the southern part of Belgium.
Material and methods
Farm sample
To analyse the relationships between input self-sufficiency,
structure, economic and environmental sustainability, we
used data derived from two regional farm accounting data-
bases (that is, Agricultural Economic Analysis Department
and Walloon Breeders Association). These databases mainly
included socio-economic and inputs-related data such as
amounts of animal feed or mineral fertilisers.
From these databases, we first selected a sample of 478
specialised dairy farms. Farms were considered as specialised
according to the definition of the European typology: in these
farms, at least 66% of the total standard gross margin was
originated from dairy cattle (European Commission, 2012).
On the basis of a diversity analysis, 80 farms were excluded
from this set because they were less specialised in milk
production (e.g. they had secondary cash crop or meat pro-
duction activities) or they highly differed from the main farm
groups identified (Lebacq
et al.
, unpublished results). In order
to avoid confusing the effect of input self-sufficiency with
that of organic production method, we considered separately
conventional and organic dairy farms: 335 conventional dairy
farms were used in the core of this study; 30 organic dairy
farms were managed independently and characterised at the
end of the result section; and 33 farms for which the infor-
mation –organic or conventional –was not known were
excluded from the analysis.
Data were available for 2008 and 2009. The year 2008
was considered as the reference year, and 2009 was used to
follow the evolution of economic results in a situation where
the milk price drops sharply. Indeed, 2009 was characterised
in Europe by an average decrease in the milk price of 24%,
compared with 2008 (European Commission, 2014).
Selection of an indicator of input self-sufficiency
Various indicators could be used as proxy to assess input self-
sufficiency of dairy farms at farm scale –that is, to assess the
extent to which the farms use small or large amounts of off-
farm inputs. We considered the following two selection cri-
teria: (1) the indicator should be measurable from the data
available in our two databases; and (2) the indicator should
not focus only on one input, such as concentrate feed, but
include several ones. In addition, indicators expressed per
1000 l of milk were not suitable for our farm sample. Indeed,
even if the farms were specialised in dairy production, some
of them also had minor secondary activities such as beef,
pork, poultry or crop production. Available data did not allow
the inputs to be correctly allocated between these different
activities. Therefore, indicators expressed per unit of product
were not used. From these criteria, we first selected the
indicator of economic dependence, which was calculated as
follows: the sum of the variable costs of animal (e.g. feed,
Input self-sufficiency of specialised dairy farms
545

veterinary products) and crop (e.g. seeds, fertilisers, pesti-
cides) production –excluding contract working –and the
fixed costs of electricity and energy (i.e. fuel, lubricants and
other energy sources) use, divided by the total gross product
excluding subsidies (Raveau, 2011). In order to facilitate the
interpretation of the indicator, we calculated an indicator
of
economic autonomy
(EA) –that is, 1 –economic depen-
dence. The higher this indicator value, the lesser the farm
uses inputs and the more the farm is self-sufficient regarding
these inputs.
EA classes and comparison among classes
In 2008, the 335 conventional farms had an average EA of
62 ± 8%
1
. In order to compare farm structure (i.e. land use,
scale and intensity of production), economic and environ-
mental performances according to their EA, these farms were
categorised into four classes. The classes were defined from
the following quartiles of EA in our conventional sample:
57% (quartile 0.25), 62% (median) and 68% (quartile 0.75).
The classes were called
Auto−−
(with the lowest degree of
EA),
Auto−
,
Auto
+and
Auto
++ (with the highest degree
of EA).
First, these classes were characterised and compared for
several structural indicators, using the Kruskal–Wallis and
multiple comparison tests. Owing to the characteristics of the
farm sample, this characterisation was performed regardless
of productive and pedoclimatic constraints: our farm sample
only included specialised dairy farms, and 93% of them were
located in two agro-ecological regions specialised in dairy
production (
Région herbagère liégeois
e and
Haute-Ardenne
),
in which the agricultural area (AA) is mainly covered by
permanent grassland and forage crops.
In order to compare the sustainability of EA classes, we
used the canonical discriminant analysis. This method allows
the differences among groups of individuals (here, the
EA classes) to be characterised through simultaneously con-
sidering several quantitative variables measured on these
individuals (here, the sustainability indicators) (Cruz-Castillo
et al.
, 1994). To perform this analysis, we used a set of
sustainability indicators (Supplementary Table S1) selected
according to the process described by Lebacq
et al.
(2013). In
order to identify differences among classes according to
sustainability performance, we excluded from this set all
indicators linked to farm structure, such as stocking rate,
permanent grassland area and economic specialisation. As
a result, 10 environmental indicators and nine economic
indicators were introduced as variables in the discriminant
analysis (Supplementary Table S1).
Third, we studied how the farms reacted to the decrease in
milk price of 2009, according to their level of EA. We centred
this analysis on the farm income per work unit, because it
represents a key aspect for maintaining farms in the agri-
cultural landscape. Farm income is the difference between
the gross operating surplus –that is, the total gross product
(including subsidies) minus variable costs, fixed costs, salary
and farm renting –and the financing costs and depreciation.
As financing costs and depreciation were not computed in
the same way in our two databases, we used the gross
operating surplus per familial work unit as proxy for the farm
income per work unit. In order to study the evolution of this
indicator between 2008 and 2009, we calculated the varia-
tion of gross operating surplus per familial work unit
between 2008 and 2009 and compared this variation among
classes through the Kruskal–Wallis and multiple comparison
tests. To help understand how the farmers dealt with this
crisis, we also calculated and compared among classes
average variations of variable costs, gross product, milk
production and input use.
Characterisation of organic farms
As organic farms are part of the agricultural landscape, we
aimed to explore the relationship between organic farming and
EA. Organic farms had an average EA of 69 ± 9%, similar to the
average of
Auto
++ (
P
=0.5). Moreover, 63% of organic dairy
farms had an EA >68% –that is, the border value defining the
class
Auto
++. Therefore, we compared the organic farms with
the class
Auto
++ for various structural, environmental and
economic (including the variation of gross operating surplus
per familial work unit between 2008 and 2009) indicators
through the Kruskal–Wallis tests.
Results
Characterisation of EA classes
Structural features of each class are shown in Table 1
2
. All
classes were similar in terms of workforce, AA, herd size,
share of heifers and total milk production. Regarding
production intensity,
Auto
++ had lower milk production per
hectare and stocking rate than
Auto−
, and lower milk pro-
duction per cow compared with
Auto−
and
Auto−−
. All
classes were forage based, with identical proportions of
forage area in the AA. However, the proportions of grassland
and maize differed among classes:
Auto
++ had a higher
proportion of permanent grassland in the AA, compared with
Auto−
and
Auto−−
, and a lower proportion of maize,
compared with
Auto−−
.
From an economic point of view, the four classes had
similar gross product, despite a slight increase in the price
for the milk delivered to the dairy according to the EA level.
Auto
++ had costs related to electricity use, animal and crop
production significantly lower than
Auto−
and
Auto−−
,
whereas
Auto
+showed intermediate average values.
The difference between extreme classes
Auto
++ and
Auto−−
was particularly large for costs of animal produc-
tion (44%). Indeed,
Auto
++ used significantly less
dairy cow concentrates. This class also bought less forage
than
Auto−−
3
. Variations of crop production costs among
1
Mean ± standard deviation.
2
Residual standard errors are provided in Supplementary Tables S2, S3 and S4.
3
This result should be interpreted with caution due to the presence of zero
values in each class.
Auto
++ included the highest proportion of zero values:
30% against 23%, 10% and 14% for classes
Auto
+,
Auto
−and
Auto
−−,
respectively.
Lebacq, Baret and Stilmant
546

classes were not associated with significant differences in
terms of nitrogen fertiliser use.
Auto
++ had energy costs
smaller than
Auto−
, but did not compensate it by a greater
use of contract workers.
Canonical discriminant analysis
The canonical discriminant analysis identified three canonical
variables. All were significant –that is, the canonical corre-
lation coefficients were significantly different from 0. The
first canonical variable had an eigen value >1 and explained
95.7% of the total between-class variance, against only
2.3% and 2.0% for the second and third variables, respec-
tively (details of the analysis are provided in Supplementary
Table S5). It means that the differences among the four
classes were important mainly in one direction. Conse-
quently, we did not consider the second and third canonical
variables in the description of the results.
The first canonical variable was interpreted from the cor-
relations with the initial variables –that is, environmental
and economic indicators (Table 2). The first canonical
variable was positively correlated mainly with the financial
dependence (i.e. the ratio annuities/gross operating surplus),
veterinary costs, energy consumption per hectare and nitro-
gen surplus per hectare (for details on the indicators, see
Supplementary Table S1). On the other hand, the first cano-
nical variable was negatively correlated mainly with the
economic efficiency (i.e. the ratio gross operating surplus/
gross product including subsidies), capital efficiency (i.e. the
ratio value added/total capital excluding land), gross margin
per hectare and gross operating surplus per familial work
unit. In other words, this variable represented the environ-
mental and economic performances of the farms: low values
of the first canonical variable corresponded to farms with low
environmental impact and good economic results.
Conventional farms and means of EA classes were plotted
on the first two canonical variables (Figure 1). As expected,
the differences among classes occurred mainly along the first
axis. Figure 1 highlights a gradient of EA along the first
canonical variable: classes with higher EA degree had sig-
nificantly lower values on this variable. On the basis of the
Table 1
Mean structural characteristics of organic and conventional dairy farms according to their degree of economic autonomy (2008)
Unit
Auto−−
1
(
n
=78)
Auto−
(
n
=80)
Auto
+
(
n
=96)
Auto
++
(
n
=81)
Organic
(
n
=30)
Workforce Annual work unit
2
1.6
a
1.6
a
1.5
a
1.6
a
1.6
Farm scale and intensity
Agricultural area Ha 59
a
61
a
57
a
58
a
68*
Herd size Dairy cows 71
a
80
a
71
a
72
a
56*
Stocking rate LU/ha
3
forage area 2.7
a,b
2.9
b
2.7
a,b
2.6
a
1.8*
Share of heifers % of cows 32
a
32
a
30
a
31
a
28
Total milk production l 477 115
a
539 629
a
462 599
a
454 547
a
305 249*
Milk yield per cow l/cow 6693
b
6666
b
6479
a,b
6218
a
5473*
Milk yield per hectare l/ha agricultural area 8201
a,b
8890
b
8230
a,b
7832
a
4592*
Land use
Forage area % of agricultural area 98
a
98
a
98
a
99
a
95*
Permanent grasslands % of agricultural area 86
a
86
a
89
a,b
93
b
87
Maize % of agricultural area 10
b
9
a,b
8
a,b
6
a
1*
Milk price €/l 0.31
a
0.32
a,b
0.33
b
0.34
c
0.42*
Gross product (without subsidies) €177 165
a
204 199
a
179 431
a
185 930
a
151 155
Costs included in EA
Animal production €65 362
c
61 725
c
46 891
b
36 349
a
33 948
Crop production €11 372
b
11 097
b
9090
a,b
8620
a
5295*
Electricity €4565
b
4594
b
3726
a,b
3579
a
3211
Other energy sources €4686
a,b
4567
b
3785
a,b
3620
a
4704*
Input use
Dairy cow concentrates kg/1000 l milk 248
c
221
b,c
209
b
171
a
166
Concentrate autonomy
4
%2
a
4
a
2
a
2
a
13*
Forage purchase kg dry matter/LU 3918
a
1012
a
3133
a,b
1913
b
1978
Mineral fertilisers kg N/ha agricultural area 96
a
94
a
86
a
87
a
16*
Use of contract work €/ha agricultural area 202
a,b
200
b
174
a,b
156
a
119
LU=bovine livestock units; EA=economic autonomy.
a,b,c
Mean values within a row with different superscripts differ significantly at
P
<0.05.
1
Auto−−
: dairy farms with an EA <57%;
Auto−
: between 57% and 62%;
Auto
+: between 62% and 68%; and
Auto
++:>68%.
2
Annual work units include familial and salaried work units. For familial workforce, 1 work unit is a farmer (or the spouse) who works full time on the farm. For salaried
workforce, the value of one work unit corresponds to 1800 h/year.
3
Livestock units were calculated from regional coefficients based on animal dietary needs.
4
Proportion of livestock concentrates (in kg) produced on the farm.
*Significant differences between the mean values of organic farms and
Auto
++ farms at
P
<0.05.
Input self-sufficiency of specialised dairy farms
547

interpretation of the first canonical variable, more autono-
mous classes were, therefore, characterised by better envir-
onmental performance –that is, lower nitrogen surplus
per hectare, energy consumption per hectare, veterinary
costs and higher nitrogen efficiency –and higher economic
performance –that is, higher economic efficiency, capital
efficiency, gross operating surplus per familial work unit,
gross margin per hectare and lower financial dependence.
These observations were globally confirmed by comparing
the mean values among classes (Table 3).
Economic impact of the sharp decrease in milk price of 2009
In average, the milk price decreased by 19 ± 9% in our
conventional farm sample between 2008 and 2009. As a
consequence, the farm income per familial work unit
decreased by 9 ± 29%, with a considerable variability among
farms. This variability was also found within EA classes.
Nevertheless, the average income variation significantly
differed between the class
Auto−−
and the other three
classes. The income of
Auto−−
was favourably impacted
with a slight average increase, whereas the other classes
experienced an average income decrease. This may be partly
explained by the following two aspects: the variation of
variable costs and the variation of the gross product
(Table 4). First, the average variable costs decreased more
sharply for
Auto−−
. This variation was mainly related to
the variation of animal production costs, as reflected by the
high correlation between both indicators (0.99, Pearson test
P
<0.001). The variation of concentrate use and concentrate
autonomy was identical across classes. However, performing
t
tests
4
for each class showed that
Auto−−
significantly
decreased the use of dairy cow concentrates (
P
<0.001),
whereas the average variation of concentrate use was not
significantly different from zero in the other classes. Second,
the decrease in the average gross product of
Auto−−
was
less marked than for other classes due to a lower average
decrease in the milk price. Despite these two characteristics,
the class
Auto−−
had, in 2009, the lowest average farm
income per familial work unit (Table 4).
Comparison between organic and
Auto++
conventional
dairy farms
Compared with the class
Auto
++, organic farms employed
similar workforce for a smaller dairy herd and lower total
production. They were more extensive systems with lower
milk yield and stocking rate. They had a wider AA, char-
acterised by a low proportion of maize. They obtained a
significantly higher milk price, allowing them to achieve a
similar gross product from smaller milk volumes. Organic
farms used similar amounts of dairy cow concentrates but
produced a larger proportion of the concentrates on the
farm. They also used no mineral fertilisers
5
(Table 1).
Regarding sustainability performance, organic farms had
economic results similar to those of
Auto
++ –that is,
income per work unit and economic efficiency –but a lower
environmental impact –that is, energy consumption and
nitrogen surplus per hectare (Table 3). In 2009, the organic
milk price was less affected by the crisis than the conven-
tional price. As a result, the average income per familial work
Table 2
Total canonical structure: correlations between the economic
and environmental indicators, and the three canonical variables
Sustainability indicators
1
Can1 Can2 Can3
Economic
Gross operating surplus
per familial work unit
−0.45 −0.43 0.19
Gross margin per hectare −0.48 −0.26 0.19
Capital efficiency −0.61 −0.17 −0.20
Economic efficiency −0.87 −0.06 −0.17
Importance of subsidies 0.20 0.13 −0.44
Financial dependence 0.43 0.21 0.29
Capital per familial work unit 0.09 −0.33 0.35
Concentrate feed autonomy −0.01 −0.27 0.21
Direct sale of milk −0.08 0.26 −0.14
Environmental
Pesticide costs per hectare 0.22 −0.35 −0.03
Soil link rate 0.12 −0.44 0.30
Sprayed area 0.02 −0.18 −0.18
Phosphorus fertilisation per hectare 0.09 0.04 0.20
Potassium fertilisation per hectare 0.09 −0.31 −0.37
Area without mineral nitrogen 0.08 0.11 −0.24
Energy consumption per hectare 0.35 −0.33 0.24
Nitrogen surplus per hectare 0.33 −0.26 0.22
Nitrogen efficiency −0.28 0.29 −0.07
Veterinary costs per cow 0.43 0.07 −0.22
1
For definitions of indicators, please refer to Supplementary Table S1.
Figure 1 Plotting of conventional dairy farms and means of economic
autonomy classes on the first two canonical variables. ■Means
of economic autonomy classes. ×Farms belonging to the class
Auto
++;+
Auto
+;□
Auto−
;Δ
Auto
−−.
4
Normality assumption was tested using a quantile–quantile plot. When the
plot was close to linear, the distribution of the indicator was considered as close
to normal (MathWorks, 2014).
5
The mean value is positive because the sample includes farms in transition
towards organic farming. In 2010, these farms were recorded as organic.
Lebacq, Baret and Stilmant
548

unit of organic farms decreased less sharply compared with
Auto
++ (Table 4).
Discussion and perspectives
Specialised and intensive livestock farming systems were
developed from an industrial paradigm focussing on practice
simplification and standardisation, as well as intensive use of
inputs (Kirschenmann, 2007). As such systems have been
called into question, increasing input self-sufficiency may
constitute a key principle to improve sustainability of live-
stock farming systems (Dumont
et al.
, 2013). In this study,
we used the indicator of EA as proxy to assess input self-
sufficiency of dairy farms. Comparison among EA classes
shows inefficiencies of some farms in the use of technical
inputs, and therefore underlines the possibility of reducing
input use, especially animal feeding, without decreasing the
total milk production.
Efficient use of inputs, structural characteristics and
sustainability performance
In our case study, reducing input use is achieved by class
Auto
++ without involving a larger production of con-
centrates on the farm. As found by Guerci
et al.
(2013), it is
Table 3
Mean economic and environmental results of organic and conventional dairy farms according to their degree of economic autonomy (2008)
Indicators
1
Unit
Auto−−
2
(
n
=78)
Auto−
(
n
=80)
Auto
+
(
n
=96)
Auto
++
(
n
=81)
Organic
(
n
=30)
Economic
Gross operating surplus €/familial work unit
3
48 999
a
71 313
b
72 072
b
82 028
b
72 650
Gross margin €/ha of agricultural area 1676
a
2084
b
2128
b,c
2381
c
1651*
Capital efficiency % 43
a
57
b
65
b,c
73
c
66*
Economic efficiency % 43
a
53
b
60
c
68
d
71
Financial dependence % 65
b
49
b
37
a
31
a
31
Environment
Energy consumption MJ/ha of agricultural area 27 478
c
26 564
b,c
23 144
a,b
19 934
a
10 172*
Nitrogen surplus kg N/ha of agricultural area 152
c
144
b,c
122
a,b
102
a
27*
Nitrogen efficiency % 31
a
32
a,b
35
b,c
40
c
71*
Veterinary costs €/cow 115
c
92
b
89
b
69
a
80
a,b,c,d
Mean values within a row with different superscripts differ significantly at
P
<0.05.
1
For detailed definitions, please refer to Supplementary Table S1.
2
Auto
−−: dairy farms with an economic autonomy <57%;
Auto−
: between 57% and 62%;
Auto
+: between 62% and 68%; and
Auto
++:>68%.
3
One familial work unit is a farmer (or the spouse) who works full time on the farm.
*Significant differences between the mean values of organic farms and
Auto
++ farms at
P
<0.05.
Table 4
Average economic impact of the milk price crisis of 2009 on organic and conventional dairy farms according to their degree of economic
autonomy
Unit
Auto
−−
1
(
n
=78)
Auto−
(
n
=80)
Auto
+
(
n
=96)
Auto
++
(
n
=81)
Organic
(
n
=30)
Variation between 2008 and 2009
2
Gross operating surplus per FWU
3
%9
a
−14
b
−14
b
−16
b
−2*
Gross product per FWU % −3
b
−10
a,b
−9
a,b
−11
a
−2*
Milk price % −17
a
−18
a
−20
a,b
−22
b
−15*
Milk production (l) % 7
a
6
a
6
a
8
a
9
Variable costs per FWU % −14
b
−11
a,b
−8
a
−5
a
−1
Animal production % −12
b
−10
a,b
−7
a,b
−4
a
−2
Crop production % −17
b
−11
a,b
−1
a,b
−2
a
38
Dairy cow concentrates % −8
a
−4
a
−1
a
−2
a
−12
Concentrate autonomy
4
%2
a
−1
a
2
a
−1
a
2
Mineral fertilisers kg N/ha agricultural area 1
a
11
a
9
a
16
a
−1
Gross operating surplus per FWU in 2009 €/familial work unit 49 789
a
60 407
b
61 594
b
68 098
b
67 753
FWU =familial work unit.
a,b
Mean values within a row with different superscripts differ significantly at
P
<0.05.
1
Auto−−
: dairy farms with an economic autonomy <57%;
Auto −
: between 57% and 62%;
Auto
+: between 62% and 68%; and
Auto
++:>68%.
2
Variation was calculated as follows: 100 ×((Value of 2009 −Value of 2008)/Value of 2008), except for the concentrate autonomy and mineral fertilisers for which the
variation was calculated as: (Value of 2009 −Value of 2008), because of the presence of 0 values.
3
One FWU is a farmer (or the spouse) who works full time on the farm.
4
Proportion of livestock concentrates (in kg) produced on the farm.
*Significant differences between the mean values of organic farms and
Auto
++ farms at
P
<0.05.
Input self-sufficiency of specialised dairy farms
549

associated with higher proportion of permanent grassland
and less maize in the AA. Indeed, permanent grasslands
require less mineral fertilisers and crop protection products
than crops (Raveau, 2011). However, this result differs from
other studies underlining the benefit of combining crop and
livestock production to reduce input use (Ryschawy
et al.
,
2012). We assume that this is related to pedoclimatic
characteristics of the study area that are poorly suited to
crop production. As a result, the AA is covered mainly by
permanent grassland that constitutes the main resource for
livestock (through grazing or mowing). Growing maize in this
area is often associated with the quest for high production
levels involving an intensive use of inputs and explaining the
higher costs of livestock production.
Concerning the environmental and economic sustain-
ability, our results show that increasing the EA level allows
the farms to have better environmental and economic
results. From an environmental point of view, several studies
have reported that farms using fewer concentrates and
mineral fertilisers have lower nitrogen surplus and energy
consumption (Hansen
et al.
, 2001; Paccard
et al.
, 2003;
Meul
et al.
, 2012). In our case study, this relationship was
established whatever the unit in which the indicators were
expressed: per hectare or per unit of product. We found, for a
subsample of 205 farms fully specialised in dairy production,
a gradual decrease in the average energy consumption
per 1000 l when the EA level increased (data not shown).
The higher use of inputs of
Auto−
and
Auto−−
does not
involve a similar increase in milk production, and therefore
results in lower nitrogen efficiency and higher energy con-
sumption per 1000 l.
From an economic point of view, the good economic
results of farms with a high EA degree can be explained by
three aspects. First, using fewer inputs for an equivalent level
of milk production allows the farms to reduce their variable
costs without affecting the gross product, and therefore have
a positive effect on their income. Second, in our case study,
Auto
++ farms received in 2008 a milk price slightly higher
than other farms. Higher prices could be explained by the
production of higher quality milk –for instance, in terms of
protein content or number of somatic cells, or by the delivery
of milk to a specific dairy paying higher price. Third, the use
of EA as indicator to categorise the farms into four classes
plays a role in these economic results. Indeed, EA is also
an indicator of economic efficiency. It explains the strong
correlation between economic efficiency, capital efficiency
and the first canonical variable. Nevertheless, we observed
that the first canonical variable was also correlated with
other economic indicators such as gross operating surplus
per familial work unit and financial dependence.
Inefficiencies in input use observed in our farm sample,
especially in terms of animal feeding, could be related to
path dependency on past agricultural policies and evolution
trajectories. Since the Second World War, agricultural sub-
sidies have been based on the output level, which pushed the
farmers to consider the gross yield as their main production
objective (Vanloqueren and Baret, 2008). In addition, the
increase in labour productivity and the process of practice
standardisation have encouraged farmers to simplify their
management practices and distribute concentrate all year
round to enable high and steady production, without taking
the quality of the forage available into account (Veysset
et al.
, 2014). Including higher proportions of forage in the
diet constitutes a possible path to reduce the use of con-
centrates without decreasing the output level. It could be
achieved through the optimisation of forage and grassland
management, and by avoiding losses during grazing, har-
vesting, preservation and feeding (Meul
et al.
, 2012). Havet
et al.
(2014) mentioned as an example the use of grassland
calendars with a visual assessment of the grass height to
optimise the management of grass quality during rotational
grazing. They also refer to research on leader and follower
grazing systems aiming to decrease the grazing of refusals by
dairy cows and assign them to animals with low dietary
requirements. However, despite better performance of more
autonomous farms, breaking away from existing routines
needs specific management skills and requires social influ-
ences (e.g. objectives of extension services and organisation
of the dairy industry) to be overcome (Meul
et al.
, 2012;
Veysset
et al.
, 2014).
Evolution in a changing context
Input self-sufficiency has often been considered as interest-
ing to promote farm sustainability, faced with an increase in
energy and input costs (López-Ridaura
et al.
, 2002; Bernués
et al.
, 2011; Ripoll-Bosch
et al.
, 2012). In addition, Raveau
(2011) found that when product prices increased, as in 2007,
autonomous farms had more latitude for increasing their
production, through increasing the use of inputs. They could,
therefore, benefit more from these increased prices. As far
as we know, no paper dealt with the role of input self-
sufficiency in a context of decrease in milk price, as observed
between 2008 and 2009. Our results underline that
Auto−−
farms are less affected by the decrease in milk price. First,
they have greater leeway for reducing the variable costs by
decreasing the use of concentrates without decreasing
strongly the milk production. A strategy of input substitution –
that is, the use of less expensive input –may also have been
implemented simultaneously; however, we did not have the
appropriate data to test this assumption. However, according
to a perception survey, farmers and consultants would
rather envisage rationalising concentrate purchase and
optimising forage and grass production to decrease feed
costs (Association Wallonne de l’Elevage, 2012). Second, the
milk crisis involved a levelling out of prices among classes in
2009, explaining the sharper price decrease for
Auto
++.
Despite these evolutions, the farms having little autonomy
kept in average lower gross operating surplus per familial
work unit in 2009.
Organic farming and EA
It was already stated that input self-sufficiency is a crucial
aspect to guarantee the economic viability of organic farm-
ing systems due to the high price of organic inputs, especially
Lebacq, Baret and Stilmant
550

concentrates (Lherm and Benoit, 2003). As established in our
study, input self-sufficiency in organic farms is achieved
through the greater use of on-farm resources and the non use
of mineral fertilisers and pesticides, thereby leading to a low
environmental impact per hectare (Hansen
et al.
, 2001;
Paccard
et al.
, 2003). Our brief comparison between organic
and conventional
Auto
++ farms shows that organic farming
provides economic benefits in a situation of milk price
decrease. In 2009, organic farms kept a stable income per
work unit because of a smaller milk price decrease. However,
although high EA levels in conventional farms do not
involve strong structural adaptations, the organic production
method is associated with extensive practices –that is, low
yields and stocking rate.
Limitations and perspectives
Our results provide new insights about the benefitofincreasing
input self-sufficiency in the context of a transition towards
more sustainable farming systems. In this respect, however,
our work shows several limitations and should be broadened in
several respects. First, the EA indicator has two main draw-
backs with respect to its ability to assess input self-sufficiency.
On the one hand, this indicator is based on economic values
and therefore varies according to input and milk prices. These
variations may skew the estimation of relative amounts of
inputs used by the farms. However, input prices could be
considered to be homogeneous among conventional farms, as
they have access to the same economic market and resources.
In addition, the impact of milk price variations on the EA
indicator was found to be negligible for conventional farms
(data not shown). On the other hand, EA constitutes an
indicator of economic efficiency. Consequently, the use of this
indicator could lead to consider economically efficient farms as
those that are input self-sufficient whereas they use a lot of
inputs, and vice-versa (Raveau, 2011). However, we found
that, for a subsample of 205 farms fully specialised in dairy
production, EA was highly correlated with the variable costs
per 1000 l (−0.77, Pearson test
P
<0.001). Therefore, farms
with a higher EA degree used fewer inputs for the same milk
production and similar input prices.
Second, we performed an analysis based on the comparison
among farms for a specific case study. Focussing on specialised
farms did not allow us to study the interest in diversifying
farming activities to increase input self-sufficiency. In fact,
input self-sufficiency is usually associated with the diversifica-
tion of farm activities in order to perform synergies and
exchanges (Vilain, 2008). For instance, a mixed farm could
increase its self-sufficiency by using manure for the crops or by
producing animal feedstuff on the farm (Bonny, 2010; Bell and
Moore, 2012). On the other hand, although this work high-
lighted interesting findings in the scope of farm sustainability, it
provides information about one type of production within a
specific agro-ecological context. Our results cannot be directly
applied to other contexts. This analysis should consequently be
broadened to explore other types of production –for example,
more diversified farming systems –within various agro-
ecological contexts. It would provide reference values of EA to
the farmers to compare their farm with farms having similar
structural constraints and opportunities. In such a bench-
marking process, it is crucial to identify practical ways to
improve EA in a specificcontext.
We assessed input self-sufficiency using a farm-level
proxy. The farm scale is the main management level and
economic unit, at which decisions, strategic choices and
technical actions are performed by the farmer to produce
goods and services (Lebacq
et al.
, 2013; Botreau
et al.
,
2014). At this level, such actions could allow the farmer to
improve the economic and environmental sustainability of
the farm. Like other sustainability attributes, input self-
sufficiency could, however, be assessed at higher levels –for
example, local, regional or national. Such assessments would
consider synergies and exchanges between farms located
in the same territory, such as the transfer of forage or
manure between crop and livestock enterprises (Bell and
Moore, 2012).
This study constitutes a first approach to analyse the
impact of input self-sufficiency on the evolution of farm
sustainability. The use of 2-year data did not allow us to take
the inter-annual variability of data into account. Moreover,
the use of farm accounting data involves possible overlaps of
processes –for example, variation of prices and volumes –
and contextual changes –for example, decrease in milk
price, increase in input price and impact of climate events.
Consequently, to further investigate this dynamic approach,
long-term studies should be performed through farm net-
work monitoring over a long period. In this context, more
detailed data should be collected to enable fine-grained
analyses in terms of processes and contextual changes, and
to take social aspects –for example, labour, into account.
Such long-term studies could explore the extent to which
input self-sufficiency supports farm development and
constitutes the core for long-term action (Havet
et al.
, 2014).
Conclusion
Input self-sufficiency is usually considered as a key aspect to
promote sustainable farming systems. In this study, we used
the EA indicator to assess input self-sufficiency of specialised
dairy farms located in the grassland area. We showed that
high EA degrees in conventional farms could be favoured by
high proportions of permanent grassland in the AA. Owing to
an efficient input use (especially in terms of animal feeding),
the most autonomous farms combine a low environmental
impact with good economic results. In the context of
milk price decrease, the least autonomous conventional
farms have greater leeway for reducing input use, without
decreasing milk production, and thereby maintaining a
stable income level. Despite this latitude, these farms have a
lower income compared with more autonomous ones. In this
context, organic farms have a stable income because of a
slighter decrease in the organic milk price. This dynamic
approach should be further investigated through long-term
studies in order to consider inter-annual variability and
various types of contextual changes.
Input self-sufficiency of specialised dairy farms
551

Acknowledgements
The first author is a recipient of a PhD grant financed by the
‘Fonds pour la formation à la Recherche dans l’Industrie et
dans l’Agriculture’(FRIA). The authors thank the Agricultural
Economic Analysis Department (DAEA) and the Walloon Bree-
ders Association (AWE) for making their data available, as well
as three anonymous reviewers for useful comments on earlier
versions of the manuscript.
Supplementary material
To view supplementary material for this article, please visit
http://dx.doi.org/10.1017/S1751731114002845
References
Association Wallonne de l’Elevage 2012. Perception survey of the dairy farmers
and their consultants. Optimir Interreg IVB. Retrieved July 31, 2013, from http://
www.optimir.eu
Astigarraga L and Ingrand S 2011. Production flexibility in extensive beef
farming systems. Ecology and Society 16, 1–13. Retrieved September 5, 2012,
from http://www.ecologyandsociety.org/vol16/iss1/art7/
Bell LW and Moore AD 2012. Integrated crop-livestock systems in Australian
agriculture: trends, drivers and implications. Agricultural Systems 111, 1–12.
Bernués A, Ruiz R, Olaizola A, Villalba D and Casasús I 2011. Sustainability of
pasture-based livestock farming systems in the European Mediterranean con-
text: synergies and trade-offs. Livestock Science 139, 44–57.
Bonny S 2010. L’intensification écologique de l’agriculture: voie et défis. In
Innovation and sustainable development in agriculture and food 2010 (ed.
E Coudel, H Devautour, C Soulard and B Hubert), pp. 1–11. Cirad, Inra,
Montpellier Supagro, Montpellier, France.
Botreau R, Farruggia A, Martin B, Pomiès D and Dumont B 2014. Towards an
agroecological assessment of dairy systems: proposal for a set of criteria suited
to mountain farming. Animal 8, 1349–1360.
Cruz-Castillo JG, Ganeshanandam S, MacKay BR, Lawes GS, Lawoko CRO and
Woolley DJ 1994. Applications of canonical discriminant analysis in horticultural
research. HortScience 29, 1115–1119.
Darnhofer I 2010. Strategies of family farms to strengthen their resilience.
Environmental Policy and Governance 20, 212–222.
Dumont B, Fortun-Lamothe L, Jouven M, Thomas M and Tichit M 2013.
Prospects from agroecology and industrial ecology for animal production in the
21st century. Animal 7, 1028–1043.
European Commission 2012. Agriculture, farm accounting data network, sample
selection. Retrieved April 6, 2012, from http://ec.europa.eu/agriculture/rica/
methodology2_en.cfm
European Commission 2014. Agriculture and rural development, European milk
market observatory. Retrieved June 25, 2014, from http://ec.europa.eu/
agriculture/milk-market-observatory/index_en.htm
Guerci M, Knudsen MT, Bava L, Zucali M, Schönbach P and Kristensen T
2013. Parameters affecting the environmental impact of a range of dairy
farming systems in Denmark, Germany and Italy. Journal of Cleaner Production
54, 133–141.
Hansen B, Alrøe HF and Kristensen ES 2001. Approaches to assess the
environmental impact of organic farming with particular regard to Denmark.
Agriculture, Ecosystems and Environment 83, 11–26.
Havet A, Coquil X, Fiorelli JL, Gibon A, Martel G, Roche B, Ryschawy J, Schaller N
and Dedieu B 2014. Review of livestock farmer adaptations to increase forages
in crop rotations in western France. Agriculture, Ecosystems and Environment
190, 120–127.
Kirschenmann FL 2007. Potential for a new generation of biodiversity in
agroecosystems of the future. Agronomy Journal 99, 373–376.
Lebacq T, Baret PV and Stilmant D 2013. Sustainability indicators for livestock
farming. A review. Agronomy for Sustainable Development 33, 311–327.
Lherm M and Benoit M 2003. L’autonomie de l’alimentation des systèmes
’élevage allaitant: évaluation et impacts économiques. Fourrages 176, 411–424.
López-Ridaura S, Masera O and Astier M 2002. Evaluating the sustainability of
complex socio-environmental systems. The MESMIS framework. Ecological
Indicators 2, 135–148.
MathWorks 2014. Quantile–quantile plot. Retrieved September 2, 2014, from
http://www.mathworks.nl/help/stats/qqplot.html
Meul M, Passel SV, Fremaut D and Haesaert G 2012. Higher sustainability
performance of intensive grazing versus zero-grazing dairy systems. Agronomy
for Sustainable Development 32, 629–638.
Paccard P, Capitain M and Farruggia A 2003. Autonomie alimentaire et bilans
minéraux des élevages bovins laitiers selon les systèmes de production.
Fourrages 174, 243–257.
Raveau A 2011. Critère d’autonomie et comportement des exploitations agri-
coles face au choc économique de 2007. Commissariat général au développe-
ment durable. Retrieved April 16, 2013, from http://www.developpement-
durable.gouv.fr/IMG/pdf/ED46.pdf
Ripoll-Bosch R, Díez-Unquera B, Ruiz R, Villalba D, Molina E, JoyM, Olaizola A and
Bernués A 2012. An integrated sustainability assessment of mediterranean sheep
farms with different degrees of intensification. Agricultural Systems 105, 46–56.
Ruiz R, Santamaria P, Arandia A, Del Hierro O, Icaran C, Intxaurrandieta JM,
Lopez E, Mangado JM, Nafarrate L and Pinto M 2011. Incorporating social and
environmental indicators in technical and economic advisory programmes in
livestock farming. In Economic, social and environmental sustainability in
sheep and goat production systems (ed. A Bernués, JP Boutonnet, I Casasús,
M Chentouf, D Gabiña, M Joy, A López-Francos, P Morand-Fehr and F Pacheco),
pp. 9–15. CIHEAM, Zaragoza, Spain.
Ryschawy J, Choisis N, Choisis JP, Joannon A and Gibon A 2012. Mixed crop-
livestock systems: an economic and environmental-friendly way of farming?
Animal 6, 1722–1730.
Ten Napel J, van der Veen A, Oosting S and Koerkamp P 2011. A conceptual
approach to design livestock production systems for robustness to enhance
sustainability. Livestock Science 139, 150–160.
Thomassen MA, Dolman MA, van Calker KJ and de Boer IJM 2009. Relating
life cycle assessment indicators to gross value added for Dutch dairy farms.
Ecological Economics 68, 2278–2284.
Vanloqueren G and Baret PV 2008. Why are ecological, low-input, multi-
resistant wheat cultivars slow to develop commercially? A Belgian agricultural
‘lock-in’case study. Ecological Economics 66, 436–446.
Veysset P, Lherm M, Bébin D and Roulenc M 2014. Mixed crop-livestock farming
systems: a sustainable way to produce beef? Commercial farms results,
questions and perspectives. Animal 8, 1218–1228.
Vilain L 2008. La méthode IDEA: indicateurs de durabilité des exploitations
agricoles. Educagri Editions, Dijon, France.
Lebacq, Baret and Stilmant
552