j o ur na l ho me page: www.elsev ier .com/ locate / landusepol
n investigation into the type of farmer who chose to participate inural Environment Protection Scheme (REPS) and the role of
nstitutional change in influencing scheme effectiveness
eraldine Murphya,∗, Stephen Hynesa,b, Eithne Murphya, Cathal O’Donoghuec
Department of Economics, National University of Ireland, Galway, IrelandSocio Economic and Marine Research Unit (SEMRU), National University of Ireland, IrelandRural Economy and Development Programme, Teagasc, Athenry, Co., Galway, Ireland
r t i c l e i n f o
rticle history:eceived 13 July 2013eceived in revised form 11 February 2014ccepted 14 February 2014
eywords:armer participation
a b s t r a c t
This paper examines the voluntary aspect of the Rural Environment Protection Scheme (REPS) in Irelandby modelling the type of farmer who chose to participate in the agri-environment programme from 1995to 2010. The impact of changing scheme payment rates and organic nitrogen restrictions on schemeuptake are also examined. In order to examine some of the heterogeneity of variable influence on par-ticipation across the different phases of REPS, separate models for each of the four phases were ran (onefor a reference year in each phase). Results from the random effects model show that the type of farmer
who was most likely to participate in REPS over time had an extensive farm system, low income andspent more hours working on-farm than their non-REPS counterparts. Single year logit model resultswere compared to the random effects panel logit model for the entire sample period. The results suggestthat the individual year models do a better job in demonstrating how farmers responded to contractualchanges in the scheme over time (in terms of their participation decision) compared to the panel model.
The Rural Environmental Protection Scheme (REPS) was createdn response to regulation (EEC) 2078 in 1994. It was an agri-nvironmental scheme (AES) available to every farmer in Irelandn a voluntary basis until 2009. As it was a voluntary scheme,he effectiveness of REPS was largely assessed by the number ofarmers who chose to join it. Consequently, scheme targets wereften defined in terms of uptake rates (Rath, 2002). Although Irisharming essentially consists of mixed livestock production, there isignificant heterogeneity in farm types across the country (Feehannd O’Connor, 2009). Hence, the effectiveness of REPS not onlyepended on the number of participants in the scheme, but also onhat type of farmer chose to join. For example, Hynes and Garvey
2009) have already shown that, from 1995 to 2005, REPS farmers
ere significantly more likely to be sheep farmers, followed by cat-
le farmers, than any other system type. The environmental impactf sheep or cattle farmers implementing REPS objectives on their
holdings is expected to be different to the environmental impactof, say, dairy farmers implementing REPS objectives on their hold-ings. Hence this paper aims to provide a description of what typeof farmer was most likely to have participated in REPS from 1995to 2010 and to discuss how these findings may have influencedscheme effectiveness during this period.
REPS had broad environmental objectives, which can be dividedinto those that tackled issues relating to biodiversity conserva-tion and those that were aimed at pollution abatement. This paperfocusses on the effect of farmer type on the attainment of pollutionabatement objectives. In REPS, pollution abatement strategies wereprimarily concerned with reducing the quantity of chemicals used,or produced, on participating farms by imposing application, oroutput, limits for chemicals on participating farms. Pollution abate-ment was then achieved in one of two ways. For farmers who usedor produced quantities of chemicals that were above the thresh-old limit, maintenance of the status quo was required. For farmerswhose levels of use or production were above the threshold limit,adoption meant they were obliged to reduce their level of chemical
use or production if they wished to participate in REPS (Emersonand Gillmor, 1999; DAFF, 2004, 2007). For the greatest levels ofscheme effectiveness to be achieved in relation to pollution abate-ment, it would have been more beneficial to have farmers joining
he scheme who needed to reduce the amount of chemicals theysed or produced on entry.
This paper focuses specifically on whether farmers appeared toroduce status quo, or reduced, levels of organic nitrogen as a con-equence of joining REPS. Hynes and Garvey (2009) believed thatheir finding indicating that REPS farmers were most likely to beheep or cattle farmers was a consequence of the restriction onrganic nitrogen production in the scheme. This is because, tradi-ionally in Ireland, cattle and sheep farms are extensive (those whoenerate less than 170 kg/ha), whereas dairy and tillage are inten-ive (Emerson and Gillmor, 1999). This result indicates that, in thearlier phases of REPS, the scheme suffered from “adverse selec-ion bias” with regard to this particular restriction: namely, thosehose participation was most likely to result in the greatest out-ut changes with regard to pollution abatement were disinclinedo join.
The objectives of REPS were applied at farm level through 11easures. There were four phases of REPS from 1994 to 2009
nd, while the topics of these 11 measures did not change acrosshases, certain institutional details regarding how they were to bepplied on individual farms did. There were two types of institu-ional changes to the contract that are expected to have impacted onhe type of farmer who joined REPS, and hence its effectiveness at
eeting its pollution abatement objectives. The first was changeso the payment rates (the majority of which were increases) andhe second was changes to the threshold limits for organic nitrogenroduction across phases. Both types of changes were intended to
mprove scheme effectiveness by attracting a greater number, and aifferent type, of farmer(s) to the scheme. Variations in farmer type
n a representative year for each phase of the REPS are discussedn relation to institutional changes to payment rates and thresh-ld limits for organic nitrogen production across phases. This ischieved using logit models of participation in the reference years999 (REPS I), 2003 (REPS II), 2006 (REPS III) and 2010 (REPS IV).he results of these logit models are then compared with a panelata model of participation for the entire time period, 1995–2010.
REPS provides an interesting case study for an investigation intohe impacts of contractual changes on participation behaviour inn AES for two reasons. Firstly, REPS was universally available toll Irish farmers and, secondly, during the period that it existed,o other AES was available to Irish farmers. Consequently, thereere few restrictions on who could join and, unlike the review
f institutional effects on participation behaviour carried out byeerlings and Polman (2009), there is no need to account for thempact of competing voluntary AESs on farmers’ preferences. Withhis in mind, the analysis in this paper should help to inform futurenternational agri-environmental policy, in addition to Irish policy.
Section “Related literature” provides a review of literatureelated to this study. Section “Policy background” provides anverview of EU and national policies that influenced Irish agricul-ure, and potentially the REPS participation decision, from 1994o 2010. Section “Estimation model and framework” describes theconomic theory underlying farmers’ participation decisions inEPS. It also shows the logit models that are used in this papero examine the types of farmer who are most likely to participaten the scheme. The NFS data used in this paper are discussed inection “Data”. Section “Results and discussion” presents and dis-usses the findings from the models of REPS participation Finally,ection “Conclusions” provides concluding remarks.
elated literature
A number of economic and sociological studies have attemptedo qualitatively and quantitatively describe farmers’ participationecisions in AESs. They have identified a number of demographic,
licy 39 (2014) 199–210
socio-economic and farm specific factors that significantly influ-ence whether a farmer chooses to join or not (Willcock et al., 1999;Dupraz et al., 2003; Ma et al., 2012). Certain authors have foundthat farm size is positively (Wilson, 1997; Lynch and Lovell, 2003)associated with the likelihood of participating in a scheme, whereasothers have found its influence is inconclusive (Wynn et al., 2001;Dupraz et al., 2003; Wossink and van Wenum, 2003). These resultsare often linked to whether the AES being studied pays farmersper hectare or not and whether there are limits on the number ofhectares that farmers are permitted to enter into the scheme.
Morris et al. (2000) find “a fear of loss of control” is a major deter-rent to participation for a number of farmers. This fear may begreater for farmers who are obligated to enter ever larger tractsof land into the scheme. Similarly, the duration of a contract maybe unattractive to farmers because it obliges them to manage theirfarms according to the specifics of the AES for a fixed length of time.In fact, Peerlings and Polman (2009) evaluate five competing landallocation contracts across Europe and find that contract length isa particularly contentious issue for potential participants.
A farm-specific variable that is almost always negatively corre-lated with participation in an AES is productivity. Examples includeindividuals’ reliance on on-farm income for household support(Defrancesco et al., 2008), high productivity potential (Dupraz et al.,2003) and productive soil types (Hynes and Garvey, 2009). This isbecause AESs often require some degree of extensification fromfarmers, the aim of which is to create habitats or reduce pollu-tion. In this case, the opportunity costs of extensification for thosewho are intensive are, clearly, higher than the opportunity costsfor extensive farmers. Wynn et al. (2001) describe it as the fit of thefarm to the scheme.
A final aspect of the impact of contractual change on farmerparticipation behaviour that needs mentioning is that farmers whohave previous experience with an AES are more likely to join acontemporary scheme than those who have not (Wossink and vanWenum, 2003). This may be because they no longer fear loss of con-trol and are comfortable with the concept of participating. It mayalso be a cyclical effect whereby the fit of their farms to a new AES isbetter as a consequence of participating in an earlier one. Whateverthe reasons, this finding suggests that retaining the same generalstructure for an AES over time is a logical decision for policy makersto make, especially if uptake rates are deemed ‘sufficient’.
Many authors argue that the participation studies describedabove do not fully capture the factors that influence AES imple-mentation. To address this problem, these authors tend to takea more qualitative approach towards the evaluation of voluntaryAESs. The role of advice providers and other non-governmentalbodies (Morris, 2004), farmers’ perceptions of farming (Wilson,1997; Willcock et al., 1999), farmers’ attitudes towards the AESin question (Burton et al., 2008) and farmers’ opinions of theconservation needs of their farms (Herzon and Mikk, 2007) haveall been shown to influence implementation outcomes for AESs.Oftentimes, factors such as farmers’ perceptions and attitudestowards farming and the environment vary spatially, meaning theirimpact on farmers’ responses to different institutional environ-ments need to be considered separately for different countries orregions (Mettepenningen et al., 2013).1
1 For further discussion on the qualitative approach towards the evaluation ofvoluntary AESs the interested reader is advised to read the review conducted byBurton et al. (2008).
se Po
P
E
awcatidDaifocftiiftn
pppoTfi
pfiw2osmsMpaita
dtfAc9tctcwtbNa(eh
G. Murphy et al. / Land U
olicy background
nvironmental and agricultural policies
A number of changes were made to both environmental andgricultural policies in Ireland from 1994 to 2010 (many of whichere made at EU level). Both environmental and agricultural poli-
ies directly influence how individuals’ manage their farms andre therefore expected to have had an impact on farmers’ par-icipation decisions in REPS. Two influential policies that werentroduced during the period studied in this paper were the intro-uction of the Single Farm Payment (SFP) scheme and the Nitratesirective (EEC 91/676). These were ratified in Ireland in 2005nd 2006, respectively. Adherence to the conditions containedn the SFP and the Nitrates Directive are both obligatory (if aarmer wishes to receive government subsidies), yet a numberf the measures contained in both policies are similar to thoseontained in the voluntary REPS scheme. This is particularly trueor the Statutory Management Requirements (SMRs) contained inhe SFP (DAFM, 2012a). The introduction of these two policiess expected to have reduced the opportunity costs of participat-ng in REPS for Irish farmers in general because they held everyarmer in the country more accountable for the negative externali-ies they emitted, regardless of whether they participated in REPS orot.
A second way that the SFP is expected to have influenced farmerarticipation behaviour in REPS was through the complete decou-ling of direct payments from production. For example, decoupledayments are expected to have reduced farmers’ temptation toverproduce agricultural goods (O’Donoghue and Howley, 2012).his meant that the opportunity costs associated with the extensi-cation requirements under REPS decreased.
In addition, farmers’ main source of on-farm income, marketrices, became increasingly volatile from 1994 to 2010. This was,rstly, a consequence of the removal of price support mechanisms,hich began with the introduction the MacSharry reforms (EEC
078) in 1992. Secondly, price volatility increased as a consequencef a move to simplify European agricultural policy, which was firstuggested at the World Trade Organisation’s Uruguay Round Agree-ent on Agriculture in 1994 (Hennessy and Thorne, 2006). This
uggestion in part led to the introduction of the single Commonarket Organisation in 2007, the aim of which was to make Euro-
ean agriculture more accessible to traders outside the EU (O’Neillnd Hanrahan, 2012). In contrast to increased volatility in on-farmncomes, REPS payments were a stable source of income, meaningheir attractiveness to farmers, particularly those who were riskverse, may have augmented over time.
Finally, changes in environmental policies aimed at conservingesignated sites are expected to have influenced farmers’ participa-ion decisions in REPS. At the beginning of this study period, areas ofarmers’ holdings may have been designated as a Special Protectionrea (SPA) under the Birds Directive (EEC 79/409) or as a Spe-ial Area of Conservation (SAC) under the Habitats Directive (EEC2/43). Farmers with designated areas on their land were obligedo adhere to management plans that often restricted their agri-ultural opportunities. These individuals are therefore expectedo have associated participation in REPS with lower opportunityosts than other farmers because their agricultural opportunitiesere limited regardless of their participation status. This observa-
ion is relevant to a 17-year examination of farmers’ participationehaviour in REPS because, in 2000, waterways of importance andatural Heritage Areas (NHAs) were added to the list of designated
reas under the Water Framework Directive and the Irish WildlifeAmendment) Act respectively. Hence, the number of Irish farm-rs with strict management plans for designated areas on theiroldings at this time increased.
licy 39 (2014) 199–210 201
The four phases of REPS
This section is concerned with identifying the institutional dif-ferences between the four phases of REPS. It also outlines howeach phase was perceived by the farming community in gen-eral, information for which has been gleaned from contemporaryliterature and newspaper reports. The contemporary general opin-ion of each phase is relevant to this study because, as withthe likes of organic farming adoption, AES adoption is expectedto be influenced by the information that is made available tofarmers (Lapple and Van Rensburg, 2011). This includes popularopinion.
Two features of the REPS I contract that have been identifiedas major deterrents to participation in REPS were the costs oflivestock housing required under Measure 1, which were consid-ered excessive for particularly small farmers, and a restriction ongrowth regulators, which discouraged tillage farmers from joining.Payment rates for this phase were D151/ha for up to 40 ha, i.e., themaximum amount of land that they could receive payments for was40 hectares. When REPS I was introduced in 1994, Irish farmers hadno previous experience of an AES. Uptake was slow to begin withand initially followed trends that mirrored that of new technol-ogy adoption in Ireland; namely, farmers who are more inclinedtowards making risky decisions were the first to opt to join thescheme (Emerson and Gillmor, 1999). However, rates of enrolmentgradually accelerated and, by 1999, REPS I had acquired its targetof 45,000 adopters (Rath, 2002). Until changes were made to theLess Favoured Areas scheme in 1995 and 1996, the number of dis-advantaged farmers who received decoupled payments specific totheir needs remained low (DAFM, 2013). REPS I filled this niche formany individuals and was viewed primarily as income support fordisadvantaged farmers by some (McEvoy, 1999).
The 11 measures contained in the REPS II contract were simi-lar to those in REPS I in every way except that farmers were nowpermitted to use growth regulators on cereals and the educationcomponent of the scheme was made compulsory. An extra tier wasadded to the payment structure, raising rates for up to 20 ha toD165/ha in an attempt to attract smaller farmers to the scheme.Farmers continued to receive D151/ha for 21–40 ha. Contemporarynewspaper reports show that REPS II was not well received by thefarming community. Reasons given for its unpopularity include thatpayment and inspection rates were, respectively, too low and toohigh (Maguire, 2003). One report highlighted farmers’ concernsthat a lack of scientific evidence showing REPS was successfullyprotecting biodiversity at farm level was creating concerns aboutfuture funding from Brussels (Maguire, 2002). Target uptake ratesfor the second phase were set at 70,000 but by 2002, they still hadnot been met (Rath, 2000, Rath, 2002).
REPS III was introduced as an amendment to REPS II in July2004. The primary change that was made to the measures con-tained in REPS III was the introduction of mandatory BUs. Thesewere intended to further change farm management practices onREPS farms with the intention of improving biodiversity levels. Par-ticipants were obliged to choose two BUs from a list of 16 in thethird phase. Payment rates increased to D200/ha up to 20 ha andD175/ha from 21 to 40 ha. In addition, two extra tiers were addedto the payment structure: D70/ha from 41 to 55 ha and D10/ha forall remaining land. Contemporary newspaper reports wrote pos-itively about this phase of the scheme, which was endorsed bythe Irish Farmers’ Association (IFA) before it was even introduced(McDonagh, 2003). In fact, articles relay a sense of urgency withregard to participation, advising farmers to join the scheme before
it is too late (Farragher, 2004) and culminating in a rush to enrolwhen the deadline for applications was announced in September2006 (Kerryman, 2006). Target uptake rates for REPS III were morethan met and it was deemed a success.
2 se Po
atif2efstm2sisac2
E
eecpa
U
U
gtofRcciohda
Y
p
Y
wpc
Y
wpbiig
P
B
most likely to participate change to complement the differentper-hectare payment rates in each phase. For this reason, farm sizeis represented as farm size categories in the four cross-sectional
2 A likelihood test was used on the pooled dataset and confirmed that durationeffects, or correlation between farmers’ participation decisions in time t and t − 1,
02 G. Murphy et al. / Land U
Adopters of REPS IV were obliged to choose at least two BUs from list of 24. However, the most significant change that is expectedo have influenced farmers’ perceptions of REPS IV is that relat-ng to organic nitrogen production. A provision was made allowingarmers with derogations under the Nitrates Directive to use up to50 kg/ha of organic Nitrogen (Hynes et al., 2008). Intensive farm-rs are expected to have found REPS IV more attractive than theormer three phases as a consequence of this change. The paymenttructure in REPS IV was the same as that of REPS III but most ofhe payment rates for the final phase were greater than for the for-
er three. Farmers were offered D234/ha up to 20 ha, D205 from1 to 40 ha and D82/ha from 41 to 55 ha. The sudden closure in theummer of 2009 of REPS IV to all new participants led to a surgen applications for the scheme. Farmers had already suffered sub-tantial losses in earnings from agricultural commodities in 2008nd 2009 as a consequence of extreme weather patterns and wereoncerned with the loss of other income sources (Connolly et al.,010).
stimation model and framework
The theoretical framework used to interpret the results of thisstimation exercise is a standard neoclassical one. When farm-rs make the REPS participation decision, they are assumed toompare the amount of utility that they expect to gain from non-articipation, UN, with the amount from participation, UR, whichre represented in the following two equations:
N(Nit, 0; Zit) (1a)
R(Pit + Nit − Cit, Eit; Zit) (1b)
At any given time t, by not participating in REPS, farmer i willain utility from farm income, Nit, outputs from which are condi-ional on a vector of farm and farmer characteristics, Zit. Elementsf Zit may also directly impact on farmers’ utility levels (Eq. (1a)). Ifarmer i does participate in REPS at time t, he will gain utility fromEPS payments, Pit as well as Nit but may also face opportunityosts, Cit, of revenue lost when farmland is altered under the REPSontract (Equation 1b). Eit is the effort associated with participat-ng in REPS. Total income Pit + Nit − Cit and Eit are both conditionaln the contents of Zit in Eq. (1b). Farmers are expected to associateigher income and lower effort levels with additional utility. Theecision function for farmer i at time t can be given as (Chambersnd Foster, 1983):
Although the value of Yit is not directly observed, a discretearticipation indicator is given by:
∗it = {0, if Yit > 0; 1, otherwise} (3)
here 1 represents participation in REPS and 0 indicates non-articipation. The decision function that the farmer evaluates whenontemplating joining the scheme can be rewritten as:
here Xit is a vector that gathers observed determinants of Yit, is aarameter vector and εit is a random component. A binary logit cane used to look at what influences farmers’ decisions to participate
n REPS, meaning the probability that farmer i chooses to participaten REPS at time t, or of the dependent variable being equal to 1, isiven as (Hensher et al., 2006):
r(Yit = 1) = 11 + exp(−ˇXit)
(5)
However, in this form, any interpretation of is cumbersome.y adjusting Eq. (5) to read as the log of the odds of participating in
licy 39 (2014) 199–210
REPS divided by the odds of not participating in REPS, the depend-ent variable reduces to a simple linear function of the explanatoryvariables:
log(
Pr(Yit = 1)1 − Pr(Yit = 1)
)= ˇ′Xit (6)
The ˇ’ represents a change in the probability of being in REPSrelative to the probability of not being in REPS associated with a unitchange in the independent variable. In this scenario, exp(ˇ′)is calledan odds ratio. The coefficients in the next section are expressed inthis way. An estimated coefficient greater (less) than 1, indicatesthat farmers are more (less) likely to participate in REPS when thereis a positive change in the explanatory variable than not. So, forexample, if a coefficient is equal to 3, that means that farmers arethree times as likely to participate in REPS than not, but if it is equalto 0.33, then they are only a third as likely.
Two aspects of the REPS participation decision are addressedin this paper using two specifications of the binary logit model.The first is what type of farmer is most likely to be found in REPSfor the duration of the scheme for which a random effects logit isused.2 This model accounts for the existence of unobserved indi-vidual effects, which exist as a consequence of individuals’ uniquedecision-making processes and therefore do not change across themultiple observations for each individual. A downside to using thismodel is that it does not allow for the existence of fixed effects,or correlations between individual effects and Xit. To compensatefor this, fixed effects are accounted for in this paper by includinga variable for average farm income across the entire period in themodel. This is a version of a Mundlak–Chamberlain random effectmodel (Mundlak, 1978) similar to that used by Hynes and Garvey(2009). It is given as:
Y∗it = ˛i + ˇ′Xit + εit (7a)
where
˛i = Iiı + vit (7b)
In Eq. (7a), Y∗it
is the indicator variable denoting whether farmer iparticipates in REPS at time t, ˛i is the individual farmer effect, Xit isa vector of explanatory variables and εit is a vector of unobservablefactors affecting whether farmer i decides to participate in REPS attime t or not. Eq. (7b) shows that ˛i is represented by mean income,I, and other unobserved influences, vit. For this reason, a variablefor I is included in Xit.
Path dependence, or duration effects, are accounted for byincluding lagged variables in Xit, indicating whether farmer i par-ticipated in REPS in the previous year or in the previous phase. Inaddition, Xit contains variables indicating which REPS phase farm-ers were participating in at time t to allow for variations in the REPScontract across phases.
To compare with the panel model above, and to gain a deeperunderstanding of participation in each phase of the scheme, theREPS participation decision is also examined using four individuallogits on NFS cross-sectional datasets from 1999, 2003, 2006 and2010, representing REPS I, REPS II, REPS III and REPS IV, respec-tively. Of particular interest to this study is whether the farm sizes
exist. Use of a fixed effects model to account for correlations between unobservedindividual effects and Xit was too restrictive for the unbalanced 16-year dataset andresulted in the loss of important information. This is due to the fact that many of theexplanatory variables vary little across the panel for individual farmers, meaningthese variables drop out of the model if a fixed effects specification is employed.
se Policy 39 (2014) 199–210 203
leottkf
D
vFctpdoaafihp
dlodoevscdd
deialfyi
Thm
wRpw
u
aiotoota
Table 1Summary statistics for the entire study period (1995–2010).
Variable NFS All NFS REPSn = 19,306 n = 5871
EconomicFarm income (D1000 s) (excl. REPS payment) 15.56 12.23Gross outputsa per ha (D/ha) 1304 1271
Farm levelSoil 1 (%) 0.47 0.44Soil 2 (%) 0.41 0.42Soil 3 (%) 0.12 0.14Farm size (ha) 36.02 36.57Farm size less than 20 ha (%) 0.37 0.26Farm size 21–40 ha (%) 0.33 0.41Farm size 41–55 ha (%) 0.13 0.17Farm size greater than 55 ha (%) 0.17 0.15Organic N (kg/ha) 102.52 90.82Dairy (%) 0.31 0.19Cattle (%) 0.46 0.50Tillage (%) 0.06 0.07Sheep (%) 0.15 0.23
DurationREPS previous phase 0.15 0.40REPS previous year 0.30 0.77
Source: NFS 1995–2010.Monetary values have been adjusted to the year 2000 using the Consumer PriceIndex (CPI). Soil type variables sum to 1, farm size categories variables sum to 1 and
G. Murphy et al. / Land U
ogit models (in the panel model, farm size is continuous). Thestimated models also allow us to examine whether restrictions onrganic nitrogen production were particularly strong deterrentso participation (thereby leading to adverse selection bias) acrosshe different phases. To investigate this question in greater detail,ernel density plots of organic nitrogen production on non-REPSarms are produced and discussed in the results section.
ata
The data used in this paper come from the National Farm Sur-ey (NFS). The NFS is collected annually by the Irish Agriculture andood Development Authority, Teagasc, as part of its data-collectionommitments to the Farm Accountancy Data Network (FADN) ofhe EU. The aim of FADN is to analyse the income and economicerformance of agricultural holdings in all EU member states. NFSatasets for 1995–2010, inclusive, are used to investigate the typef farmer who was most likely to participate in REPS over time. Inddition, NFS datasets for the individual years of 1999, 2003, 2006nd 2010 are used to estimate the likelihood of finding differentarmers in REPS I, II, III and IV, respectively. This is because, accord-ng to the Department of Agriculture, Food and Marine (DAFM), theighest percentages of REPS farmers representing each of the fourhases participated in these years.3
The NFS data include information on individuals’ farm level,emographic and duration characteristics. Included under the farm
evel characteristics heading are variables for soil type farm size,rganic nitrogen production and system type. The soil types areesignated by Teagasc. Each farm in the NFS is classified into onef three major soil groups according to their use range (Hennessyt al., 2011).4 Farm size is given as utilisable agricultural area. Theariable for organic nitrogen production is created from the den-ity of livestock found on each farm. System type is based on theommunity typology of agricultural classification5 and refers to theominant enterprise on each representative farm. System typesiscussed in this paper are dairy, cattle, sheep and tillage.6
Table 1 shows how values for the variables used in this studyiffered between REPS farmers and Irish farmers in general for thentire study period. Average farm income (all sources of on-farmncome except REPS payments) and productivity levels (estimateds market based gross outputs per hectare) for REPS farmers wereower from 1995 to 2010 on REPS farms than they were for thearming population generally. REPS farmers worked 80.36 h less perear (or just over 1.5 h per week) on their farms than Irish farmersn general.
The statistics for demographic variables in both samples inable 1 are similar. The average age for all NFS farmers is marginallyigher than that of REPS farmers alone. REPS farmers are slightlyore likely to be married (74%) and to have children (46%) than all
3 In other words, 1999 was the year when the highest percentage of REPS farmersere participating in REPS I, 2003 was the year when the highest percentage ofEPS farmers were participating in REPS II, 2006 was the year when the highestercentage of REPS farmers were participating in REPS III and 2010 was the yearhen the highest percentage of REPS farmers were participating in REPS IV.4 Soil type 1 has the widest use range and soil type 3 contains farms with limited
se range.5 This is a uniform classification of holding used throughout the EU.6 Using reductions in the production of organic nitrogen as an indicator for the
chievement of pollution abatement objectives on tillage farms is admittedly notdeal. This is because tillage farmers in the NFS from 1995 to 2010 used an averagef 87.46 kg/ha of synthetic nitrogen on their farms, which is significantly more thanhe average amount used by all other farm systems (3.17 kg/ha). However, only 6%f these NFS farms are classified as tillage (Table 1) and each one has livestock unitsn their farms. Hence, whilst the reduction in organic nitrogen production underhe REPS scheme is not assumed to be a primary concern for tillage farmers, it didpply to them.
system type variables sum to 1.a Market based gross outputs only.
Irish farmers from 1995 to 2010 (67% and 42% of the entire sam-ple are married and have children respectively). Consequently, interms of the demographic variables considered in this paper, REPSfarmers more strongly adhere to the profile of a farmer who is morelikely to consider long-term goals in their decision making than not(Hennessy and Rehman, 2007).
Average farm size for REPS farms differs little from that for allNFS farms (36.57 ha versus 36.02 ha). However, the dummy vari-ables indicating farm-size distributions show that a greater numberof REPS farms were from the mid-ranges of size categories than thesmallest (less than 20 ha) or the largest (greater than 55 ha) sizeranges. Average organic nitrogen production on REPS farms is lowerthan average organic nitrogen production for the entire Irish farm-ing population. This may be linked to the fact that the most to leastlikely farming system types found in all Irish farms are cattle, dairy,sheep and tillage, whereas for REPS farms the most to least likelysystem types are cattle, sheep, dairy and tillage. Hence, the secondmost predominant system type found in REPS is not dairy, whichis the most intensive system type in Ireland, but sheep, which isgenerally an extensive system type.
Table 2 shows average values for the variables used in this studyfor all Irish farms in 1999, 2003, 2006 and 2010 (these are the yearsrepresenting REPS I, REPS II, REPS III and REPS IV respectively). Par-ticipation in REPS decreases from 32% to 27% from 1999 to 2003but increases to 48% and 45% in 2006 and 2010, respectively. Thesenumbers are expected given that target uptake rates for REPS I werethe lowest of all four phases at 45,000 adopters, and that REPS II wasan unpopular phase that only succeeded in attracting 31,687 whenits target uptake rate was 70,000 (Maguire, 2003). REPS III and REPS
IV were considered more attractive phases than REPS II. They suc-ceeded in drawing a greater number of farmers to the scheme forreasons that included a reduction in opportunity costs associated
204 G. Murphy et al. / Land Use Po
Table 2Summary statistics for all Irish farms in the years representing the four REPS phases.
Variable 1999 2003 2006 2010n = 1061 n = 1207 n = 1141 n = 994
REPS participation 0.32 0.27 0.48 0.45
EconomicFarm income (D1000 s) (excl. REPS payment) 12.00 15.93 16.27 18.32Gross outputsa per ha (D/ha) 1095 1267 1383 1473
Sources: CSO (2012), NFS (1999, 2003, 2006, 2010).Monetary values have been adjusted to the year 2000 using the CPI. Soil type vari-ables sum to 1, farm size categories variables sum to 1 and system type variablessum to 1.
a Market based gross outputs only.b Average wage earned by Irish employees not in agriculture, forestry or fishing.
Includes industrial earnings; distribution and business services; banking, insuranceand building societies; public sector and construction.
watt
oeiwswideyfttiti
rs
2010 sample (91.99 kg/ha). Altogether, these findings suggest that,in 2010, REPS farms were only marginally more extensive (if even)than Irish farms in general.
7 The number of individuals in the smallest size category was 161 in 1999, 156 in2003, 154 in 2006 and 104 in 2010.
8 Statistics from the Central Statistics Office show that total heads of cattle (beefand dairy) fell from approximately 7.9–6.9 million and total heads of sheep fell from
ith joining resulting from the introduction of other agriculturalnd environmental policies between 2003 and 2006. In addition,hese phases offered farmers higher payment rates for participationhan the former two phases.
The final row of Table 2 shows values for earnings received byther sectors in the economy in each year. The amount farmersarned working on-farm was lower than that of people workingn other sectors in every instance displayed in Table 2. However,
hereas average farm income and productivity increased fromample year to sample year, the average value of other sectors’ages decreased from 2006 to 2010 (CSO, 2012). These statistics
mply that, for the former three sample years, the farming sectorid not benefit from the prosperity that affected the country in gen-ral. However, unlike other sectors, 2010 was a relatively successfulear for Irish farming (Connolly et al., 2010). Related to this is theact that average on-farm hours dropped substantially from 1999o 2003 and again from 2004 to 2006. This is likely due to the facthat farmers were supplementing their on-farm income by work-ng off the farm for significantly higher wages during a period whenhe demand for manual workers, particularly in the constructionndustry, was high.
The size of Irish farms in the NFS generally increased over theeference time period. In particular, the number of farms in themallest size category fell from 37% of the sample in 1999 to 30% in
licy 39 (2014) 199–210
2010.7 Conversely, organic nitrogen production decreases for eachsample year. Another way of viewing this is that the average num-ber of heads of cattle per hectare on Irish farms fell during thisstudy period (CSO, 2014).8 This reduction in livestock density is pre-sumed to have been largely caused by the decoupling of paymentsfrom production over time (O’Donoghue and Howley, 2012). Thedecrease in organic nitrogen production may also be a consequenceof a reduction in the number of dairy farms in Irish agricultureacross phases (from 33% of the NFS in 1999 to 22% in 2010). Thisreduction was largely driven by the establishment of higher yield-ing cows, combined with the introduction of production constraintsfrom milk quotas, over time (Dillon et al., 2008).
Table 3 shows that the highest percentage of REPS farmers whoparticipated in the previous phase of REPS was for the 2003 (REPS II)sample (81%), followed closely by 2010 (REPS IV) sample (71%). Thelikelihood of being in REPS in the previous year is almost constantacross the samples, which is expected given that all four phaseslasted for five years. The likelihood of being in the previous phaseis higher than the previous year for REPS II because a number offarmers in the sample did not go directly into REPS II from REPS I,meaning they were non-REPS farmers in 2002.
Average farm income for REPS farms was only 62.17% of theaverage farm income for all farms in 1999 (7460/12,000). Simi-larly, it was 64.77% of the total NFS dataset (10,350/15,930) in 2003.However, in 1999 and 2003, gross outputs per hectare from marketbased earnings were 95.93% (1046/1095) and 94.87% (1202/1267)of the national average, respectively. These findings suggest thatlow incomes on participating farms in REPS I and REPS II were dueto farm level attributes outside of those related to market basedearnings. For example, REPS I and REPS II farms were substantiallysmaller than the average size of all NFS farms in 1999 and 2003,respectively (Table 2 and Table 3).9
In 2006 farm incomes on REPS farms increased to 79.59%(12,950/16,270) of the entire sample. Gross outputs per hectare onREPS III farms in Table 3 are 96.38% of the national average (Table 2).The increase in farm income from previous REPS phases is likely tobe due to this increase in gross outputs per hectare (when com-pared to the national average) in 2003, as well as the increase inaverage REPS farm size. In particular, more REPS III farmers belongto the largest size category (greater than 55 hectares) than REPS Ior REPS II. It is also worth noting that the amount of time that REPSparticipants spent working on farm is lowest for REPS III, which isthe year with the highest value for other sectors’ earnings (Table 2).This may indicate that REPS farmers took greater advantage of off-farm earnings during the more prosperous years than other Irishfarmers.
In REPS IV, farm incomes increased to 90.17% (16,520/18,320)of the entire sample. Gross outputs per hectare were marginallyhigher than those for the national average (1552/1473). The aver-age size of REPS IV farms in Table 3 is almost the same as thenational average in Table 2 (39.72 ha versus 40.32 ha). In addi-tion, REPS IV farmers worked longer hours on their farms thanthe entire sample and the difference in organic nitrogen produc-tion (89.81 kg/ha) is only slightly smaller than that of the whole
7.9 to 4.8 million in Ireland between June 1999 and June 2009.9 To fully comprehend the causes of the differences in farm income, however,
requires a comparison of average direct payments and costs of production on REPSfarms with those of the entire sample, which goes beyond the scope of this study.
G. Murphy et al. / Land Use Policy 39 (2014) 199–210 205
Table 3Summary statistics for REPS farms in the years representing the four REPS phases.
Variable 1999 (REPS I) 2003 (REPS II) 2006 (REPS III) 2010 (REPS IV)n = 335 n = 335 n = 550 n = 458
EconomicFarm income (D1000 s) (excl. REPS payment) 7.46 10.35 12.95 16.52Gross outputsa per ha (D/ha) 1046 1202 1333 1552LabourAnnual on-farm hours 1841 1758 1693 1755
onetary values have been adjusted to the year 2000 using the CPI. Soil type variaba Market based gross outputs only.
Average organic nitrogen production on REPS farms fell acrosshe four phases and is lower for REPS farms than the NFS averagen each case. Differences between the REPS averages and the NFSverages may be due to the organic nitrogen restrictions in the REPSontract. They may also be caused by the large number of sheeparmers participating in REPS, which remained at 25% of the sam-le for REPS I, REPS II and REPS III and dipped only slightly to 22%or REPS IV. A comparison of the variable for organic nitrogen pro-uction in Tables 2 and 3 reveals that, whilst the average amountf organic nitrogen produced on all Irish farms decreased from08.85 to 91.99 kg/ha in 1999 and 2010, respectively, the quan-ities on REPS farms marginally increased from 88.07 kg/ha in 2003o 89.81 kg/ha in 2010. This finding shows that Irish farms becameess intensive (on average), or that REPS became more attractive toncreasingly intensive farmers, over time. The extent to which thishange came about because farmers changed their farm manage-ent practices to enable participation in REPS is examined in the
ext section.
esults and discussion
The results of a random effects logit model on farmer participa-ion behaviour in REPS from 1995 to 2010 are displayed in Table 4elow. The first thing to note is that the significance of the variableor average farm income indicates that an individual level effectxisted across time periods in the Mundlak–Chamberlain model.
Given the five-year duration of all REPS contracts and the num-er of similarities among the four phases, the high and significantalues for participation in the previous year as well as the previoushase displayed in Table 4 are unsurprising. It can also be seen that
EPS farmers have, ceteris paribus, lower farm income (excludingEPS payments), work longer hours on farm and have higherroductivity levels than non-REPS farmers. These findings suggesthat REPS payments are supplementing lower farm incomes. They
m to 1, farm size categories variables sum to 1 and system type variables sum to 1.
also imply that farmers are being reimbursed for the extra efforttaken to employ new management regimes on their farms as partof the scheme (such as the establishment of hedgerows or mainte-nance of stonewalls on their holdings) more so than for reductionsin production that occur as a consequence of implementing thescheme on their farms. The fact that this model shows REPSfarmers had, ceteris paribus, marginally higher gross outputs perhectare than non-REPS farmers may be explained by the actionsof extensive farmers whose land has limited production potential.The logical option for business-minded, extensive farmers whowish to optimise profits from their holdings is to maximise produc-tion on their farms (which is unlikely to breach REPS stipulations)in addition to joining REPS. This behaviour has been witnessed inIrish farmers in relation to the SFP, whereby individuals continueto produce higher than expected levels of commodities despitethe fact that the SFP is decoupled from production (Hennessy andThorne, 2006; O’Donoghue and Howley, 2012).
Demographic variables in Table 4 indicate that REPS farmers aremore likely to be young and to be married than non-participants.This profile of farmer has been linked with the type of individualwho is more actively involved in AES participation than not (Wilson,1996; Defrancesco et al., 2008). Consequently, the type of farmerswho actually joined REPS may be more likely to actively engage inAES implementation on their farms than those who did not chooseto join.
While the random effects model in Table 4 describes the typeof famer who participated in REPS from 1995 to 2010, it does nothelp to decipher how this type of famer changed across the fourphases. The high significance of the dummy variables for the REPSphases in Table 4 indicate that there was, indeed, a different effect
of each phase on farmers’ participation decisions over time. Of par-ticular interest to this paper is whether changes that were made topayment rates and permissible levels of organic nitrogen produc-tion across the four phases impacted on the type of farmer in each
206 G. Murphy et al. / Land Use Po
Table 4Results from a random effects logit on REPS participation from 1995 to 2010.
Variable Odds ratio SE
DurationREPS previous phase 0.735 (0.115)***
REPS previous year 10.848 (0.071)***
EconomicAverage income (D1000 s) 0.974 (0.005)***
Farm income (D1000 s) (excl. REPS payment) 0.958 (0.004)***
Gross outputs per ha (D/ha) 1.002 (0.001)***
LabourAnnual time working on farm (hours) 1.001 (0.001)***
Source: NFS (1995–2010).N = 16,867. Monetary values have been adjusted to the year 2000 using the CPI.Standard errors in parentheses.
a Base REPS phase is REPS I.b Base soil type is soil 3 (least productive).
cBase system type is sheep.*p < 0.1.*
p(iatitdpmti
C
emioTlw
di
in this paper (changes to payment rates and the limit on organicnitrogen production) appears to have a strong impact on REPS Iparticipation. Firstly, despite the fact that individuals did not get
11 Kernel density plots are modifications of histograms. They are smoothed plots
*p < 0.05.*** p < 0.01.
hase. In Table 4 the sign for farm size is positive. Hynes and Garvey2009) ran a similar model to look at farmer participation behaviourn REPS from 1995 to 2005 and found that farm size was negativelyssociated with participation. The differences in the findings fromhese two studies suggest that there was significant heterogene-ty with respect to the influence of farm size on participation overime. Table 4 also shows that, while REPS farmers have higher pro-uctivity levels than their non-REPS counterparts, they are likely toroduce less organic nitrogen than non-participants. This findingay be a consequence of adverse selection bias brought about by
he restrictions on organic nitrogen included in the scheme. Thesessues are investigated using the models described below.
hanges across phases
In order to examine some of the heterogeneity of variable influ-nce on participation across the different phases of REPS, separateodels for each of the four phases were ran (one for a reference year
n each phase). Table 5 below shows the results of four logit modelsn cross-sectional data representing each phase of the scheme.10
he combined direction and significance of the variables in the fourogit models are each unique. This implies that the type of farmer
ho chose to join each of the four phases was different.
10 The NFS weighting system is used for the cross-sectional logits (and kernelensity functions) but not the panel data model because the relative weights for
ndividual farms change over time.
licy 39 (2014) 199–210
As discussed previously, the two institutional effects studied inthis paper are the impact of payment rates and organic nitrogenrestrictions on scheme uptake. The impact of payment rates can beseen from the coefficients on the farm-size categories in Table 5.The coefficients for organic nitrogen production in Table 5 tell onlyhalf the story of the impact of organic nitrogen production on par-ticipation. This is because negative coefficients on this variable mayindicate that REPS farmers reduced their organic nitrogen produc-tion levels to meet the requirements of the scheme or they maybe a consequence of endogeneity in the model. Endogeneity wouldmean that participation in REPS is as much a consequence of havinglower levels of organic nitrogen as having lower levels of organicnitrogen is a consequence of participation in REPS. In other words,the coefficients for organic nitrogen in Table 5 may indicate thatindividuals with low levels of organic nitrogen production weremore likely to join the scheme in the first place than those withhigh levels (in which case REPS suffered from adverse selectionbias). To gain a better understanding of whether this is the case,histograms and kernel density plots for organic nitrogen produc-tion on non-REPS farms for the four phases of the scheme wereexamined (Fig. 1).11
The highlighted bins on the histograms in Fig. 1 indicate therestriction level for the production of organic nitrogen on REPSfarms for the first 3 phases (170 kg/ha). Ideally, the kernel den-sity plots showing organic nitrogen production levels on non-REPSfarms should be normally distributed. This would indicate that non-REPS farmers with all levels of organic nitrogen production are aslikely to join the scheme as the entire sample.12 If, however, in anyone year the 170 kg/ha threshold level specifically impacted on thelikelihood that farmers participated in REPS, there will be an obvi-ous impact of the restriction on the individual distributions in Fig. 1.In particular, if farmers who produce higher levels of organic nitro-gen are less likely to join a particular phase of REPS because they donot want to reduce their stocking rates to meet the requirementsof the scheme, there will be a decrease in the percentage of non-REPS farmers in the area of the corresponding kernel density plotbelow the threshold level of 170 kg/ha and an increase in numbersafter the threshold in Fig. 1. Such a kink in the distribution of non-REPS farmers indicates that the organic nitrogen restriction led toadverse selection bias in the scheme.
Table 5 shows that the type of farmer who participated in REPSI had poorer soils, lower farm incomes and worked fewer hourson-farm than their non-REPS counterparts. These findings implythat REPS I farmers were predominantly from areas where the landhad limited production potential and that they earned less fromon-farm production than non-REPS I farmers. In fact, a major rolethat REPS I played when it first appeared in Irish agriculture wasas income support for disadvantaged farmers (McEvoy, 1999). Thefindings in Table 5 therefore suggest that REPS I farmers did notneed to reduce their production levels to adhere to the scheme’srequirements.
Neither of the two groups of institutional changes reviewed
that show, for each selected point, the proportion of the sample that is near it. Near-ness is defined by a weighting function called the kernel function, which has thecharacteristic that the further a sample observation is from a selected point, thesmaller its received weight (Hensher et al., 2006). Specifically, the kernel densityfunction that is used to show organic nitrogen production on non-REPS farms aboveis of an Epanechnikov form. The bandwidth used in figure 4.2 is 10 kg/ha.
12 A kdensity plot for all farms in the NFS from 1995 to 2010 compared to a normaldistribution plot demonstrates that organic nitrogen use for all farms over the periodis almost perfectly normally distributed.
G. Murphy et al. / Land Use Policy 39 (2014) 199–210 207
Table 5Results of the binary logits on farmer participation in the years representing the four REPS phases.
Variable 1999 (REPS I) 2003 (REPS II) 2006 (REPS III) 2010 (REPS IV)Odds ratio (SE) Odds ratio (SE) Odds ratio (SE) Odds ratio (SE)
Sources: NFS (1999, 2003, 2006, 2010).N = 109,396 in 1999; 105,994 in 2003; 106,058 in 2006; 88,560 in 2010 (NFS frequency weights used). Monetary values have been adjusted to the year 2000 using the CPI.Standard errors in parentheses.
a Base REPS phase is REPS I.b Base soil type is soil 3 (least productive).
c
*
*
pc4l
Fph
S
Base system type is sheep.p < 0.1.*p < 0.05.*** p < 0.01.
aid for any additional hectares above 40, the most likely farm-sizeategories found in this phase are greater than 55 ha followed by1–55 ha. This result may be partially explained by the cost of
ivestock housing required under the nutrient-management plan
0.0
05.0
1.0
150
.005
.01
.015
0 100 200 300
1999
2006
Histogram
ig. 1. Histograms and kernel density plots of organic nitrogen production on non-REPS
roducing the various quantities of organic nitrogen on their farms (or the percentage of
istogram shows the bin for 170 kg/Ha organic nitrogen production level.
ource: (NFS, 1999, 2003, 2006, 2010).
in REPS, which was considered excessive for particularly smallfarmers (Emerson and Gillmor, 1999). However, the seeminglytotal disregard for the payment structure of the scheme is mostprobably due to the need for an alternative to production-based
0 100 200 300
2003
2010
Kdensi ty plot
farms for the four phases. Vertical axis shows the percentage of non-REPS farmersfarmers contained in each bin of the histogram) for each phase. The blue line in the
2 se Po
pdteiow
tou2aRo2pAR
lotmwftIVdepp
pImoaintjao
ttpalapfotril
tasmts
08 G. Murphy et al. / Land U
ayments among disadvantaged farmers. Secondly, the kernelensity plot for non-REPS I farmers in Fig. 1 shows that, althoughhe graph is skewed to the right, there is only a minimal thresholdffect (kink) around the 170 kg/ha point for REPS I farmers. Thismplies that, although the farmers who joined the first phasewned predominantly extensive farm types, the threshold itselfas not a main deterrent to participation in 1999.
Table 5 shows that REPS II farmers were 9.954 times more likelyo have been REPS I farmers than not. Therefore, a large numberf REPS I farmers participated in REPS II. Policymakers set targetptake rates at 70,000 farmers in the second phase, which was5,000 more than for REPS I. For this, policymakers provided fewdditional incentives to entice farmers to join. Payment rates underEPS II had not kept pace with inflation rates at the time and onef the primary incentives for joining REPS I had disappeared by003: namely, disadvantaged farms were now in receipt of newayments as a consequence of changes made to the Less Favouredreas Scheme in 1995 and 1996, meaning they were less reliant onEPS payments for income support.
The type of farmer who was likely to participate in REPS II hadower farm income (excluding REPS payments) but higher grossutputs per hectare and on-farm hours than their non-REPS coun-erparts. They also had more productive soil types. These results
ay be linked to the fact that the likelihood of finding dairy farmers,ho generally operate in more productive areas than, say, sheep
armers, in REPS II was higher than in REPS I. A possible reason forhe high ceteris paribus likelihood of finding dairy farmers in REPSI may have been the outbreak of foot-and-mouth disease in 2001.ollenweider et al. (2011) have shown that REPS is used by Irishairy farmers as a risk management tool, so certain dairy farm-rs may have appreciated definite sources of income from REPSayments when compared with the risk of losing income fromroduction if their herd contracted the disease.
Participants in REPS II appear to be more cognisant of theayment rates and organic nitrogen restrictions than those in REPS
. Increases in payment rates for this phase ended at 40 ha and theost likely farm-size category found in REPS II is 21–40 ha. The
dds ratio for organic nitrogen production in REPS II is negativend there is a strong threshold effect (kink) in Fig. 1. These find-ngs suggest that the negative impact of the odds ratio for organicitrogen production in Table 5 is due to extensive farmers joininghe scheme rather than being a consequence of intensive farmersoining the scheme and reducing their chemical use. Therefore, itppears as though those who joined REPS III were not the ideal typef farmer for meeting the scheme’s pollution abatement objectives.
A total of 48% of those in the NFS dataset in 2006 are REPS par-icipants (Table 2), so this phase of the scheme had the potentialo make the greatest amount of environmental change of all fourhases. They have the highest odds ratio values for the marriagend children variables of the four logits, meaning they are the mostikely participants of the REPS phases to have successor influences,nd they are also younger than their non-REPS counterparts. Thisrofile of REPS III farmers suggests that they may be the type ofarmer who views the decision to participate in REPS as some sortf long-term plan for the farm (Hennessy and Rehman, 2007). Thisype of farmer is capable of producing environmentally effectiveesults under REPS from the point of view that most of the measuresncluded in the scheme’s contract should ideally be maintained foronger than the five-years.
Payment rates in REPS III were not only greater than those forhe previous two phases but two additional tiers were added fordditional per hectare rates above 40 ha. The findings for the farm-
ize categories in the REPS III logit reflect these changes, with theost likely farm-size category being 41–55 ha followed by greater
han 55 ha. In addition, the magnitudes of the odds ratios for farm-ize categories in this logit are lower than those of the REPS I and
licy 39 (2014) 199–210
REPS II logits. This means that not only did this phase of the schemeattract a greater number of larger-sized farms to the scheme, but ithad greater variability in its farm-size categories than the previousphases.
Despite this heterogeneity in farm-size categories in REPS III,the likelihood of finding a variety of system types in REPS III is thelowest of all four phases of the scheme. These findings indicate thatthe increase in payment rates in REPS III did not introduce greatervariation to the type of farm that was most likely to be found inREPS, but merely resulted in an increase in the number of extensive,albeit larger, farms covered by the scheme. Similarly, changes to thepayment rates in REPS III did not overcome adverse selection biasas a consequence of the existence of the organic nitrogen restric-tion. The threshold effect in Fig. 1 is the strongest of all four REPSphases, meaning the larger farms that joined REPS were not reduc-ing their production of organic nitrogen to meet the requirementsof the scheme, but were maintaining them at pre-participation lev-els. This finding suggests that the type of farmer who participated inREPS III was not ideally suited to meeting the pollution abatementobjectives of the scheme.
The type of farmer who was most likely to participate in REPSIV had the smallest farm income difference from their non-REPScounterparts of all four phases. The two most likely farm-size cat-egories found on REPS IV farms are 41–55 ha followed by greaterthan 55 ha, which is presumably a consequence of the paymentstructures for the fourth phase of the scheme. The values of theodds ratios for the farm-size variables in the 2010 logit are lowerthan those in the REPS III logit, meaning variability in farm sizewas even greater for REPS farms in 2010 than in 2006. There isalso a substantial amount of variability in the type of system foundin REPS IV, which suggests that the fourth phase of the schemeattracted a wider variety of farm types, as oppose to larger sizedextensive farms. In addition, the kernel density plot in Fig. 1 hasonly a small kink in its smoothness surrounding the 170 kg/ha bin,whereas other years had a more substantial threshold effect. Thislatter finding suggests that the removal of the 170 kg/ha restrictionfor individuals with derogations in REPS IV resulted in a change inthe type of farmer who joined the fourth phase that was not evidentto the same extent in previous phases when payment rates werechanged.
Conclusions
The CAP intends to introduce further AESs in the near future(DAFM, 2012b). The success of current and future AESs is depend-ent upon what can be learned from the successes and failures of pastschemes. In the Irish case, all the required information must there-fore come from an evaluation of REPS. Results from the panel modelshow that the type of farmer who was most likely to participatein REPS over time had an extensive farm system, low income andspent more hours working on-farm than their non-REPS counter-parts. They were also younger, more likely to be married, had poorersoil types and produced lower levels of organic nitrogen than theirnon-REPS counterparts. When modelled separately (a representa-tive year for each phase of the scheme), the type of farmer whoparticipated in each of the four phases of REPS was different to thisrepresentative farmer from the random effects model. This suggestsfarmers responded to contractual changes in the scheme over time.
REPS I participants were primarily disadvantaged farmers inneed of income support. Their choice to participate in the schemeis not expected to have produced optimal outcomes in terms of
the pollution abatement objectives of the scheme. In REPS II, par-ticipants may have been using payments as a form of risk aversionafter an outbreak of foot-and-mouth disease resulted in them beingconcerned about future farm incomes. Nonetheless, this phase of
se Po
tjuttttfsc
jiHpiif
tpihRciwn
aebaeirtTE
R
B
C
C
C
C
k
D
D
D
D
G. Murphy et al. / Land U
he scheme failed to entice the required number of farmers tooin, meaning it failed to meet even the basic objective of requiredptake rates. There is evidence of adverse selection bias with regardo organic nitrogen production in REPS III, which can be attributedo the 170 kg/ha restriction contained in the contract. It is likelyhat this bias heavily influenced the low level of variability in sys-em types joining this phase. Adverse selection bias is almost absentrom REPS IV and the findings for organic nitrogen production andystem variation in 2010 were both conducive to the scheme beingapable of producing effective environmental results.
Overall, these findings show that the number of farmers whooined REPS, and the size of farms participating in the scheme,ncreased in response to larger payment rates in the latter phases.owever, the type of farmer who responded to the changes inayment rates may not have been the ideal type of farmer for meet-
ng the pollution abatement objectives of the scheme. This is anmportant consideration given that the value of REPS payments toarmers totalled D3.3 million in 2006 alone (DAFF, 2009).
An alternative to increasing payment rates is to make institu-ional changes that result in a decrease in the opportunity costs ofarticipation in the scheme for farmers. Findings from this paper
ndicate that removing restrictions on organic nitrogen productionad a superior effect on improving the type of farmer who joinedEPS across phases than changing payment rates. However, thisonclusion must be considered with caution because, in this case,ncreasing scheme effectiveness by improving the type of farmer
ho joined REPS came as a consequence of reducing the effective-ess of a scheme measure.
As georeferenced habitat data become increasingly more avail-ble to researchers, the methods used in this paper could be used tovaluate in greater detail what type of farmer participated in REPSy habitat types on the farms. The results from this paper wouldlso be greatly enhanced by using qualitative research to ask farm-rs why they behaved in the way they did with regard to REPS. Bysolating the motivations behind the decisions made by farmers inelation to REPS participation, a deeper understanding of the fac-ors driving involvement in an AES by landowners may be gleaned.his information would be useful in the design of any new AESs foruropean agriculture.
eferences
urton, R., Kuczera, C., Schwarz, G., 2008. Exploring farmers’ cultural resistance tovoluntary agri-environmental schemes. Sociologia Ruralis 48, 16–37.
hambers, R., Foster, W., 1983. Participation in the farmer-owned reserve program: adiscrete choice model. American Journal of Agricultural Economics 65, 120–124.
onnolly, L., Kinsella, A., Quinlan, G., Moran, B., 2010. NationalFarm Survey 2009. Rural Economy Research Centre (RERC) Tea-gasc, Dublin, Available at: http://www.agresearch.teagasc.ie/rerc/downloads/NFS/NFSReport09.pdf
SO, 2012. Earnings [online]. Central Statistics Office, Available at:http://www.cso.ie/en/statistics/earnings/ [accessed 16.01.12].
SO, 2014. Central Statistics Office. Statbank, Livestock [online]. http://www.cso.ie/px/pxeirestat/database/eirestat/Livestock%20and%20Farm%20Numbers/Livestock%20and%20Farm%20Numbers statbank.asp?SP=Livestoc[accessed 2014].
AFF, 2004. Terms and conditions of the Rural EnvironmentProtection (REPS) in implementation of (EC) 1257/1999 asamended by (EC) 1783/2003. Food, Department of Agricul-ture Fisheries, Dublin, Available at: http://www.agriculture.gov.ie/farmerschemespayments/ruralenvironmentprotectionschemereps/pastruralenvironmentprotectionschemereps/reps3/
AFF, 2007. Terms and Conditions of the Rural Environment ProtectionScheme (REPS) and Natura 2000 (1698/2005). Food, Department of Agri-culture Fisheries, Dublin, Available at: http://www.agriculture.gov.ie/ruralenvironment/ruralenvironmentprotectionschemereps/reps4/
AFF, 2009. REPS Factsheet. Facts and Figures 2006 [online]. Department of Agri-culture Fisheries and Food, Available at: http://www.agriculture.gov.ie/
AFM, 2012a. Cross Compliance [Online]. Department of Agricul-ture, Food and the Marine, Available at: http://agriculture.gov.ie/farmerschemespayments/crosscompliance/ [accessed 13.11.12].
licy 39 (2014) 199–210 209
DAFM, 2012b. The future shape of the Common Agricul-tural Policy [online]. Department of Agriculture, Food andthe Marine, Available at: http://www.agriculture.gov.ie/media/migration/farmingschemesandpayments/commonagriculturalpolicycap/captowards2020/Requestforsubmissionsbackgrounddocumentfinal.pdf[accessed 01.10.12].
DAFM, 2013. Categories of Disadvantaged Areas [online]. Department of Agricul-ture, Food and the Marine, Available at: http://www.agriculture.gov.ie/farmerschemespayments/singlepaymentschemedisadvantaged-areasschemebeefdataprogrammebdp/categoriesofdisadvantagedareas/[accessed 23rd September].
Defrancesco, E., Gatto, P., Runge, F., Trestini, S., 2008. Factors affecting farmers’participation in agri-environmental measures: a northern Italian perspective.Journal of Agricultural Economics 59, 114–131.
Dillon, P., Hennessy, T., Shalloo, L., Thorne, F., Horan, B., 2008. Future outlook forthe Irish dairy industry: a study of international competitiveness, influence ofinternational trade reform and requirement for change. International Journal ofDairy Technology 61, 16–29.
Dupraz, P., Vermersch, D., DeFrahan, H., Delvaux, L., 2003. The environmen-tal supply of farm households. Environmental and Resource Economics 25,171–189.
Emerson, H., Gillmor, D., 1999. The rural environment protection scheme of theRepublic of Ireland. Land Use Policy 16, 235–245.
Farragher, F., 2004. Pressure increases on Dept. to get REPS III back into operation.Connaught Tribune, 1st of January 2004.
Feehan, J., O’Connor, D., 2009. Agriculture and multifunctionality in Ireland. In:McDonagh, J., Varley, T., Shortall, S. (Eds.), A Living Countryside: The Politicsof Sustainable Development in Rural Ireland. Ashgate Publishing Ltd., Surrey,UK.
Hennessy, T., Moran, B., Kinsella, A., Quinlan, G., 2011. National Farm Survey2010. Rural Economy Research Centre (RERC) Teagasc, Dublin, Available at:http://www.teagasc.ie/publications/2011/1016/NFS10.pdf
Hennessy, T., Rehman, T., 2007. An investigation into factors affecting the occu-pational choices of nominated farm heirs in Ireland. Journal of AgriculturalEconomics 58, 61–75.
Hennessy, T., Thorne, F., 2006. How decoupled are decoupled payments? The evi-dence from Ireland. Eurochoices 4, 30–35.
Hensher, D., Rose, J., Greene, W., 2006. Applied Choice Analysis: A Primer. CambridgeUniversity Press, New York.
Herzon, I., Mikk, M., 2007. Farmers’ perceptions of biodiversity and their willingnessto enhance it through agri-environment schemes: a comparative study fromEstonia and Finland. Journal for Nature Conservation 15, 10–25.
Hynes, S., Garvey, E., 2009. Modelling farmers’ participation in an agri-environmental scheme using panel data. An application to the RuralEnvironment Protection Scheme in Ireland. Journal of Agricultural Economics60, 546–562.
Hynes, S., O’ Donoghue, C., Murphy, E., Kinsella, A., 2008. The impact of REPSparticipation on farm chemical input usage and the production of negativeexternalities. Tearmann: Irish Journal of Agri-environmental Research 6, 15–28.
Kerryman, 2006. No rush on REPS. Kerryman, 13th of September 2006.Lapple, D., Van Rensburg, T., 2011. Adoption of organic farming: are there differences
between early and late adoption? Ecological Economics, 70.Lynch, L., Lovell, S., 2003. Combining spatial and survey data to explain par-
ticipation in agricultural land preservation programs. Land Economics 79,259–276.
Maguire, D., 2002. Why farmers aren’t queuing up to join REPS 2. Irish FarmersJournal, 2nd of November.
Maguire, D., 2003. Sweeping changes sought in REPS. Irish Farmers Journal, 18th ofJanuary.
McDonagh, G., 2003. The IFA view on REPS. Irish Farmers Journal, 28th of May.McEvoy, O.,1999. Impact of REPS – Analysis from the 1998 Teagasc National Farm
Mettepenningen, E., Vendermeulen, V., Delaet, K., Van Huylenbroeck, G.,2013. Investigating the influence of the institutional organisation ofagri-environmental schemes on scheme adoption. Land Use Policy 33,20–30.
Morris, C., 2004. Networks of agri-environmental policy implementation: a casestudy of England’s Countryside Stewardship Scheme. Land Use Policy 21,177–191.
Morris, J., Mills, J., Crawford, I.M., 2000. Promoting farmer uptake of agri-environment schemes: the Countryside Stewardship Arable Options Scheme.Land Use Policy 17, 241–254.
Mundlak, Y., 1978. On the pooling of time series and cross section data. Econometrica46, 69–85.
O’Donoghue, C., Howley, P., 2012. The Single Farm Payment: a basic income forfarmers? Basic Income Studies 7, article 4.
O’Neill, S., Hanrahan, K., 2012. Decoupling of agricultural support payments: the
impact on land market participation decisions. European Review of AgriculturalEconomics 39, 639–659.
Peerlings, J., Polman, N., 2009. Farm choice between agri-environmental contractsin the European Union. Journal of Environmental Planning and Management 52,593–612.
actual and contingent participation. European Review of Agricultural Economics30, 461–485.
Wynn, G., Crabtree, B., Potts, J., 2001. Modelling farmer entry into the Environmen-
10 G. Murphy et al. / Land U
ath, F., 2000. The new REPS Scheme 2000. 2000, National REPSConference. Teagasc, Dublin, Available at: http://www.teagasc.ie/publications/reps2000/reps2000 paper01.asp
ath, F., 2002. REPS 2000 and Beyond. 2002. National REPS Conference. Teagasc,Available at: http://www.teagasc.ie/publications/2002/reps/reps2002.pdf
ollenweider, X., DiFalco, S., O’Donoghue, C., 2011. Risk Preference and VoluntaryAgri Environmental Schemes: Does Risk Aversion Explain the Uptake of theRural Environment Protection Scheme? Grantham Research Institute on ClimateChange and the Environment. London School of Economics and Political Science,
Working paper No. 48.
illcock, J., Deary, I., Edwards-Jones, G., Gibson, G., McGregor, M., Sutherland, A.,Dent, B., Morgan, O., Grieve, R., 1999. The role of attitudes and objectives infarmer decision making: business and environmentally-oriented behaviour inScotland. Journal of Agricultural Economics 50, 286–303.
licy 39 (2014) 199–210
Wilson, G., 1996. Farmer environmental attitudes and ESA participation. Geoforum27, 115–131.
Wilson, G., 1997. Factors influencing farmer participation in the Environmen-tally Sensitive Areas scheme. Journal of Environmental Management 50,67–93.
Wossink, G., van Wenum, J., 2003. Biodiversity conservation by farmers: analysis of
tally Sensitive Areas schemes in Scotland. Journal of Agricultural Economics 52,65–82.
