Modelling the Impact of Decoupling on Structural Change in the Farming Sector: integrating econometric and optimisation models

RERC Working Paper Series 06-WP-RE-01




The Rural Economy Research Centre
Working Paper Series

Working Paper 06-WP-RE-01





Modelling the Impact of Decoupling on Structural Change in
the Farming Sector: integrating econometric and
optimisation models

Thia C. Hennessy and Tahir Rehman










rerc


Irish Agriculture and Food
Rural Economy Research Centre
Development Authority
Ionad Taighde Eacnamaíochta Tuatha
For More Information on the RERC Working Paper Series
Email: CODonoghue@rerc.teagasc.ie, Web: www.tnet.teagasc.ie/rerc/

---

RERC Working Paper Series 06-WP-RE-01



Modelling the Impact of Decoupling on Structural Change in the Farming Sector:
integrating econometric and optimisation models


Abstract
This paper analyses implications of the 2003 Mid Term Review of the CAP, particularly decoupling
of direct payments from production, for Irish farming. Using the Irish FADN data base, an
integrated modelling approach involving optimisation models and econometric estimation has
been developed to analyse the continued economic viability and the changing structure of
farming. Farm level adjustments in response to policy reform are modelled to facilitate the
estimation of the effect of a policy on structural aspects of Irish farming such as the number of
farms, the proportion of full and part-time farms, the number of dairy farms, the volume of
production, the level of farm income and the viability of farming. The first step in this approach is
to develop, and solve annually, a profit maximising linear programming model for each farm
included in the FADN data set. These linear programming models use results generated by three
exogenously estimated models. These results are of three types: (i) estimates of the effect of
policy on the rate of entry and exit from farming and thus farm numbers; (ii) more ‘positive’
projections of the effect of policy on the allocation of farm labour; and, (iii) projections of
reallocation of exiting farmers’ land and milk quota, within the sector and quantitative estimates
of how policy changes might affect that reallocation. The use of the proposed modelling
approach suggests that farm numbers will decline over the next five years and, the rate of
decline wil accelerate further after decoupling relative to a continuation of Agenda 2000 policies.
Decoupling is likely to result in a more positive economic outlook for beef farming with an
increase in the number of economically viable beef farms. The number of beef farmers relying on
income from outside the farming sector will however increase. Dairy farmers will face a price cost
squeeze and that the pace of structural change in this sector of farming will accelerate due to
decoupling. Despite an increased availability of milk quota for farmers remaining in business, the
number of economically viable dairy farming businesses is set to decline.

For More Information on the RERC Working Paper Series
Email: CODonoghue@rerc.teagasc.ie, Web: www.tnet.teagasc.ie/rerc/

---

RERC Working Paper Series 06-WP-RE-01

Introduction
The Mid Term Review (MTR) of the Common Agricultural Policy (CAP) has allowed for the
decoupling of all direct payments from production from 2005 onwards; until then, most direct
payments were coupled to production, requiring farmers to produce specific products in order to
claim support. After decoupling, farmers will receive a payment regardless of production as long as
their farm land is maintained in accordance with good agricultural practices. Direct payments to
farmers have been an integral part of the CAP since the 1992 Mac Sharry reforms. Throughout the
1990s, market prices for farm produce have declined generally in line with policy while costs of
production have continued to increase. Meanwhile, direct payments increased in value, increasing
farmers’ reliance on this source of income. Furthermore, farmers adapted farming practices to
maximise their receipt of direct payments, leading to the culture of ‘farming the subsidy’. By 1997,
on cattle and tillage farms in Ireland 100 per cent of family farm income was derived from direct
payments, meaning that on average the market-based revenue was insufficient to cover total costs.
Farmers engaged in production only to receive the payments, see Figure 1.

Figure 1: Direct Payments as a Percentage of Family Farm Income on Irish farms
175
150
125
t
age 100
r
c
en
75
e
P
50
25
0
1993
1995
1997
1999
2001
2002
Dairying
Cattle Rearing
Cattle Finishing
Tillage

Source: Irish National Farm Survey, Teagasc.

For More Information on the RERC Working Paper Series
Email: CODonoghue@rerc.teagasc.ie, Web: www.tnet.teagasc.ie/rerc/

---

RERC Working Paper Series 06-WP-RE-01

The decoupling of direct payments is expected to have major ramifications for aggregate
agricultural production, farm practices and the structure of farming in Ireland. It will significantly
reduce the actual ‘coupled’ return to production; and, in some cases, the return to coupled
production will be negative. This paper presents a modelling approach developed to assess the
changes that are likely to be engendered by decoupling, in terms of the implications for the
economic viability and the structure of farming in Ireland. The paper begins by providing some
background to the economics of decoupled payments and the challenges of modelling such policy
instruments. Following on from this, the proposed modelling approach is outlined and described.
The results of the modelling exercise are presented and the paper concludes with some
recommendations for future research.

The Challenge of Modelling Decoupling and its Relationship with Structural Change
The difficulties of expanding the EU within the constraints of a limited agricultural budget, the
desire to make agriculture more market oriented and, the perceived need to formulate policies that
are defensible within the current WTO processes which have lead to pressure for reform of the
direct payment system in place for the EU farmers. It was in response to these pressures that the
Luxembourg Agreement was ratified in June 2003, making it possible to decouple all (or some)
direct payments from production.1 In Ireland, all payments are decoupled from production from
January 2005. A decoupled payment is based on the number of premiums received in a historical
reference period, paid in the form of a per hectare Single Farm Payment (SFP) the land under
farming during the reference period.

Economic theory suggests that if coupled subsidies are replaced with decoupled payments, then
production falls to a level that would exist without any subsidies. If such a situation transpires, then
production on farms making a market-based loss should fal substantially post decoupling unless

1 For further details on the partial decoupling options included in the Luxembourg Agreement see European Commission
(2003).
For More Information on the RERC Working Paper Series
Email: CODonoghue@rerc.teagasc.ie, Web: www.tnet.teagasc.ie/rerc/

---

RERC Working Paper Series 06-WP-RE-01

significant cost management or efficiency gains can be achieved. The production effects of
decoupled payments however are still somewhat of an enigma. Burfisher and Hopkins (2003) have
reviewed research on the topic to show that even fully decoupled payments have a ‘production
inducing effect’ as they affect farmers’ exposure to economic risk, their access to capital and their
future expectations. Whilst direct payments may be decoupled from production there may still be an
‘incentive effect’, which can occur if some residual production or resource use is still required to
qualify for the decoupled payment (Swinbank 2004). Although production is not necessary after the
MTR, the direct payment remains tied to land. Even if payments were not to be linked to production
at all, supply will not be so price sensitive so as to immediately fal to the free trade levels, which is
especially the case for multi-period activities such as livestock.

With or without a link to production, payment is a source of revenue for the farm household and
thus it may indirectly affect production decisions through what is referred to as a ‘wealth effect’.
Hennessy (1998) and Sckokai & Moro (2002) have explored the interaction between decoupled
payments, farmers’ risk preferences and production decisions. They conclude that if farmers’
aversion to risk declines as income increases, then an increase in wealth can induce them to take
riskier production decisions; thus, output increases compared to the situation when no decoupled
payment is made. Decoupled payments also relax the household’s capital constraint, lowering the
cost of capital to the household. According to Andersson (2004) the resulting effect is that farm
investment is likely to be greater after decoupling than in the absence of such payments. Revell and
Oglethorpe (2003) have recently explored the expectations effect, claiming that producers may
adopt a ‘safety first’ strategy and make only minimal changes to production plans in case future
payments are reassessed and again related to production or an agricultural activity. It is clear then
that even decoupled payments can influence production decisions. This paper explores some of
these issues empirical y. Whilst farmers’ risk preferences or investment plans are not modelled
explicitly, the effect of decoupling on production decisions, entry and exit decisions and the pace of
structural change in farming is explored.
For More Information on the RERC Working Paper Series
Email: CODonoghue@rerc.teagasc.ie, Web: www.tnet.teagasc.ie/rerc/

---

RERC Working Paper Series 06-WP-RE-01


Despite the long-standing interest in structural change in farming, modelling such change still
remains notoriously difficult (Garvey and Steele 1999). The processes of structural change play a
powerful role in the analysis of competitive industries in standard microeconomic text books, but as
noted by Gale (2002), there has been relatively little empirical study of the process in farming. The
available empirical models of structural change in agriculture mostly focus on the aggregate by
examining changes in the total number of farms using time-series econometric models or changes
in the numbers in various sub-sections of the population using, for example, Markov Chain models.
Such aggregate modelling approaches are often criticised for overlooking the micro dynamics of
change (Jackson-Smith 1999). Furthermore, such models do not lend themselves conveniently to
policy analysis as it is difficult to quantify the relationship between policy instruments and changes
in farm numbers.

The Markov Chain is probably the most frequently used model for analysing structural change.
Recently, non-stationary Markov Chain models have been used to project changes in the structure
of farming in response to exogenous shocks, see (Zepeda 1995; Karantininis 2001; and. Jongeneel
2002). Theoretical y, the non-stationary Markov Chain model would analyse the effect of a policy
reform and likewise, regression techniques could be used to estimate the effect of the new policy
on the probability of farms moving from one structural state to another. There are however two
main reasons why a Markov Chain model is not appropriate for the research questions addressed in
this paper. First, the limited details available in the Irish macro data it is not possible to develop a
model that allows movement between all states of structural change; that is, a matrix of transition
probabilities for all n*n cells cannot be estimated. It is therefore necessary to use a Krenz-modified
Markov Chain, which assumes that an identifiable pattern of structural change is evident; for
example, farms getting bigger, only small farms exiting and entry only through one size class. This
assumption is not tenable for Ireland, as exits from farming occur from all sizes and systems and
farms of all sizes and systems choose to transfer into part-time farming. Furthermore, given the
For More Information on the RERC Working Paper Series
Email: CODonoghue@rerc.teagasc.ie, Web: www.tnet.teagasc.ie/rerc/

---

RERC Working Paper Series 06-WP-RE-01

major policy reform under investigation, new structural states may evolve, for example the
existence of the “sofa farmer”, and the Krenz-modified Markov Chain model cannot predict
unprecedented structural states.2

The second problem in using non-stationary Markov Chains is the estimation of the transition
probabilities; the model assumes that the historical relationships between the various exogenous
variables and the transition probabilities remain constant into the future. This assumption is not
sustainable in analysing the effect of a change in intervention prices or export subsidies, that is
the policy instruments are the same and there is simply a marginal adjustment to their value.
Decoupling is an unprecedented change to policy and hence the coefficients estimated from
regression analysis on data from an Agenda 2000 type policy regime would not be appropriate
for decoupling.3 Furthermore, with decoupling new policy instruments emerge, most notably the
SFP. To analyse the effect of the SFP in a non-stationary model, it is necessary to identify a proxy
for the SFP. Identification of a suitable proxy variable, that is a source of revenue to the
household that is linked to land but not to production, is problematic. Given these difficulties, it
was decided to move away from a Markov Chain type methodology and instead to develop a
farm level model of structural change.

Methodology
In this paper the FAPRI-Ireland model is adapted to estimate the effect of decoupling on the pace
of structural change in Irish farming. The FAPRI-Ireland Partnership consists of a partial equilibrium
model of Irish agriculture, which is linked to the FAPRI EU GOLD model, and a set of farm level
models. At the aggregate level, a set of individual econometrically estimated commodity models are
linked and solve simultaneously under different policy scenarios. The farm level modelling system is

2 A sofa farmer is one who uses the farm land only to claim the decoupled payment but not to produce any tangible
agricultural output.
3 This criticism is due to Lucas (1976) who, in his seminal paper, argued that empirical models estimated under a specific
policy regime are not applicable for economic analysis under another policy regime because the parameters of an
estimated model embody the policy under which the data were generated.
For More Information on the RERC Working Paper Series
Email: CODonoghue@rerc.teagasc.ie, Web: www.tnet.teagasc.ie/rerc/

---

RERC Working Paper Series 06-WP-RE-01

comprised of a number of representative farms that are model ed using multi-period profit
maximising linear programming (LP) models. This modelling system however, does not account for
how policy may affect the pace of structural change in faming or the reallocation of resources that
may occur as a result of a policy change. In this paper the FAPRI-Ireland farm level models are
adapted to estimate the effect of decoupling on structural change in farming.

The proposed methodology involves integrating econometric and optimisation models. A profit
maximising LP framework is retained to simulate production decisions. The advantage of LP is it does
not rely on time-series data and it does not extrapolate future relationships from historical ones, and
therefore it can go beyond the realm of past observations and analyse unprecedented changes. The
disadvantages of using LP however are its normative nature and its limited scope to project
population change. To overcome these weaknesses, the LP model is supplemented with a number of
exogenously estimated models of farmer behaviour that can quantify the effects of non-pecuniary
factors on farmers’ decision-making. Three exogenous models were estimated: first, entry to and exit
from farming; second, labour allocation; and third, land and milk quota distribution. The first model
simulates the Irish farming population. The second model estimates the number of part-time farmers
and the amount of farm labour to provide the right hand side parameters for the labour constraint in
the LP models. The third model simulated the al ocation of land and milk quota; again, to provide
the right hand side parameters for the land and quota constraints in the LP models.

Modelling Entry and Exit Decisions
Many studies of entry and exit in the farming sector have concluded that age related variables are
the most significant factors (Gale 1999 and Glauben et al 2003). Gale noted that there is a common
perception that farm numbers are in decline due to accelerated exits. His research on farm numbers
in the US in the 1950s and 1960s, however, shows the decline is mostly due to a substantial drop in
new entrants concurrent with a steady rate of retirement. An age cohort analysis of the Irish data
reveals that farm numbers in Ireland are in net decline as older farmers leaving the sector exceed the
For More Information on the RERC Working Paper Series
Email: CODonoghue@rerc.teagasc.ie, Web: www.tnet.teagasc.ie/rerc/

---

RERC Working Paper Series 06-WP-RE-01

young new entrants. Hence entry and exit from farming are model ed in the context of succession
and retirement decisions. Several empirical models of retirement were developed, including early
retirement scheme and heir identification models. Due to the lack of verifiable empirical data and in
the absence of a statistically significant model, it was necessary to assume that the retirement
process is independent of the agricultural policy environment and that retirement occurs on average
at 70 years of age, as suggested by previous qualitative research (Gasson, Errington and Trantrer
1998). Better empirical data are available on the succession decisions and it is therefore possible to
quantify the factors affecting a young person’s decision to enter farming.

The decision to enter farming is modelled in the context of the nominated farm heir’s occupational
choice between farm and non-farm work (Hennessy and Rehman 2006). Drawing on the seminal
contribution by Schmidt and Strauss (1975), a model of occupational choice is developed.
Theoretically, an individual chooses his/her eventual occupation by comparing the discounted utilities
derived from all alternative occupations over the entire expected life-span of a career and, then
chooses the occupation that maximises life-time utility (Barkley 1990). The individual i is assumed to
have a subjective evaluation of each occupation type and to choose the occupation with the highest
utility index. Thus for the individual i faced with j choices, the utility of choice j is

ij
U = α + β ' ij
x
+
ij
ε







(1)
where β ' ij
x is a function of the observed attributes of the alternative, the occupational choice
and the observed characteristics of the decision-maker and ε ij , the random component,
represents the unobserved attributes of the occupations and the decision-maker. If the individual
makes the choice j = 1 then Uij is maximised from among the j utilities. The empirical model is
driven by the probability that choice j is made, that is:
Prob

( ij
U > ik
U
)
∀ k
≠ j




(2)
For More Information on the RERC Working Paper Series
Email: CODonoghue@rerc.teagasc.ie, Web: www.tnet.teagasc.ie/rerc/

---

RERC Working Paper Series 06-WP-RE-01

The above probability is estimated using the multinomial logit model (MNL). In the MNL xij
denotes the vector of variables that influence the utility associated with each occupational choice
j as perceived by each individual heir i. The probability that individual i will choose occupation j is
exp (β ' xij)
Prob(

i

chooses

j) =




(3)
∑m exp(β'xik)
k = 1
where m equals the number of occupations in the choice set. It is assumed that the nominated
farm heir is faced with three choices; full-time farming, a non-farming occupation and part-time
farming; that is, combining both farm and non-farm work.4

Using data collected by the Irish National Farm Survey (NFS) on farmers’ succession plans and
their heirs’ occupational choices the above MNL model can be estimated. Farmers participating in
the survey were questioned about their succession plans and their nominated farm heirs’ future
plans. Farmers were asked first if they had nominated an heir and subsequently about what they
expected their heir to do in future, i.e. continue the farm or not.5 The nominated heirs’
occupational choice is represented by the categorical variable CHOICE. The empirical data
suggest that part-time farming is the most common occupational choice as reported by 48 per
cent of respondents, whereas just 21 percent of farms are likely to continue on a full-time basis.
Using the MNL framework, the farm and personal characteristics that are hypothesised to affect
the succession can be tested empirically. Table 1 presents the variables included in the MNL
model.

The results of the MNL model show that an heir’s educational achievements influence all
occupational choices significantly (appendix 1). Interpreting the effect of education on the

4 Whilst there may be many non-farming occupations, they have been combined to one occupational category here as our
interest is specifically in the probability of entering farming.
5 The data on the nominated farm heirs’ occupational choices suffers from generational bias in that it reflects the current
generations’ opinions of what their heirs will do rather than the heirs’ actions or plans. However, it is the only such data
available for this study.
For More Information on the RERC Working Paper Series
Email: CODonoghue@rerc.teagasc.ie, Web: www.tnet.teagasc.ie/rerc/

---