• Agri-environment;
  • Common Agricultural Policy;
  • difference-in-difference analysis;
  • farm incomes;
  • rural environmental policy


  1. Top of page
  2. Abstract
  3. 1. Agri-environmental Policy in England
  4. 2. The Policy Evaluation Problem and Difference-In-Difference Approach with Propensity Score Matching
  5. 3. Take Up of ELS amongst Cereal Farms in Eastern England
  6. 4. Examining ELS Impact on Income
  7. 5. Discussion and Conclusions
  8. References

Applying a difference-in-difference approach with propensity score matching, we examine the impact of participation in the Entry Level Stewardship (ELS) scheme on cereal farm incomes in eastern England. We assess the extent to which impacts are related to a) the source of income affected – whether only from agricultural or total business income; b) the channel of the impact – through land use and/or labour input; and c) the level of impacts through time. In addition, we assess the appropriateness of the level of the ELS payment. We find that: a) entering the ELS scheme could negatively affect cereal farm incomes – in particular, the total business income; b) that negative impacts arise primarily in relation to the use of land resources; c) that impacts may diminish over a relatively short period of time; and d) that the ELS payment broadly compensates for losses without providing over compensation. Given the diminishing negative impact over time, the level of ELS payment might need to be reviewed in the longer term, although policy evaluation should consider the wider implications for efficiency of alternative payment levels. We also discuss some limitations of the approach and potential extensions.

1. Agri-environmental Policy in England

  1. Top of page
  2. Abstract
  3. 1. Agri-environmental Policy in England
  4. 2. The Policy Evaluation Problem and Difference-In-Difference Approach with Propensity Score Matching
  5. 3. Take Up of ELS amongst Cereal Farms in Eastern England
  6. 4. Examining ELS Impact on Income
  7. 5. Discussion and Conclusions
  8. References

Agri-environment schemes were first introduced in the European Community in the mid-1980s in response to concerns about the harmful impacts of the increasing intensity of farming coupled with the exchequer costs of dealing with surplus production under the stimulus of Common Agricultural Policy (CAP) price support (Baldock and Lowe, 1996). Schemes were initially permitted under European Structures Regulation (797/85) in 1985 but were subsequently made mandatory in European Member States in a package of measures (Council Regulation 2078/92) accompanying the MacSharry reforms of the CAP in 1992. Following the Agenda 2000 reforms, they have been included under Rural Development Regulations as part of Pillar 2 of the CAP.

Agri-environment schemes are voluntary contracts under which farmers are offered payments for adopting environmental management measures that go beyond the requirements of regulations or cross-compliance (European Commission, 2005). These cover such actions as taking land out of production, reducing livestock stocking densities, managing hedges and ditches or planting crops for wild birds. Purvis et al. (2009) estimate that there are probably in excess of 355 EU-funded agri-environment schemes, varying widely in terms of structure, scope and focus and covering three general issues: natural resources, biodiversity and landscape quality. The experience of agri-environment schemes in Europe has recently been reviewed by Hodge (2013b).

Contracts last for at least 5 years under which farmers receive annual payments. The criteria for determining the levels of payment are set out in the Rural Development Regulation (RDR), currently covering the period 2007–13, in Regulation 1698/2005. Payments should be made “in accordance with the polluter-pays principle” and should “cover only those commitments going beyond the relevant mandatory standards” in order to cover “additional costs and income foregone resulting from the commitment made. Where necessary, they may cover also transactions cost” (RDR, Article 39). Actual payment rates are determined by national governments in their Rural Development Programmes, approved by the European Commission. In practice, payments offered to farmers are determined on the basis of partial budgets based on standard farm business management data. There has, however, been very little empirical analysis of the extent to which participation in agri-environment schemes actually affects farm incomes in practice.

A number of agri-environment schemes have been operated in England, the most important being the Environmentally Sensitive Areas and Countryside Stewardship schemes operating from 1986 and 1991, respectively, until they were closed to new entrants in 2004. Since 2005 these schemes have been replaced by Environmental Stewardship, primarily in the forms of Entry Level Stewardship (ELS) and Higher Level Stewardship (HLS). ELS2 supports a wide range of basic environmental management options that aim to tackle environmental issues affecting the wider countryside, such as diffuse pollution and biodiversity, by encouraging the majority of farmers to participate. ELS is available to farmers and land managers on a voluntary basis and the ELS management options were specifically designed to be achievable by most farmers and land managers and to fit with the management of their conventionally farmed land (Defra, 2007, pp. 5–67). Points are allocated to options on the basis of the income foregone and costs. In order to receive payment, farmers select options so as to achieve the equivalent of 30 points per hectare for their whole farm area. The numbers of points associated with options are determined by Defra according to European Commission rules and the UK government deliberately set rates for Environmental Stewardship that “were generally well below 100%” of income foregone (IF). “Only where there was a significant environmental need, e.g. to meet biodiversity targets, were rates set at 100% of IF”.

There are various types of cost experienced by farmers participating in agri-environment schemes. Some requirements incur direct costs, such as cutting hedges or maintaining ditches. There are also opportunity costs in terms of production activities that are foregone, such as cropping areas lost through the introduction of buffer strips. Several of the options available in ELS involve taking land out of commercial production. Finally there are transactions costs (Falconer, 2000) in terms of the administrative requirements of learning about scheme specifications, making an application, implementing changes and monitoring and keeping records as required.

Jones and colleagues (Jones, 2005; Wallis and Jones, 2007; Grey and Jones, 2008; Harrison and Jones, 2010) have undertaken a series of case studies of the costs to individual farms of entering into agri-environment schemes. They focus primarily on ELS and find a consistent pattern of payments in excess of the levels of costs incurred. Grey and Jones (2008) studied seven farms of varying types in Gloucestershire and found that on average costs represented an estimated 27% of the value of the payments received. Harrison and Jones (2010) undertook a study of eight farms in the Lincolnshire Wolds and found a similar pattern of farms profiting from participation in the ELS, although the extent depended on the type of environmental features on the farm and whether arable land was taken out of production. However, there was considerable variation across different farms and it is difficult to generalise on the basis of these case studies.

Langton (2011) has undertaken an analysis of the links between economic performance and the introduction of agri-environmental schemes amongst a large sample of cereal farms. This finds a positive relationship between membership of schemes and the economic efficiency of the farm business, suggesting that “the relatively minor adjustments to agronomic practices required by ELS have not had too much of an adverse effect on the economics of agricultural production, and that the income received from the scheme has been sufficient to make the overall impact on the business positive” (Langton, p. 29). However, as noted in the report, it is not clear whether scheme membership may be making businesses more efficient, or whether more efficient farmers tend to join the schemes. This possibility of systematic differences between the farms that do join the scheme and those that do not is the problem of selection bias and the methodology adopted in this paper sets out to resolve this issue.

The costs of entering an agri-environment scheme arise from changes in the costs and returns of many aspects of the farm business, and surveys of individual farms cannot identify the possible extent to which changes in costs from year to year are due to entry into a scheme. Thus for instance, it would be difficult to estimate the proportion of office costs associated with ELS administration or the decline in net income due to a particular field corner being taken out of production. Similarly, these costs may also not be estimated reliably through the partial budgeting approach adopted in setting the payment levels for agri-environment schemes based on “average” farm conditions. Chabé-Ferret and Subervie (2013) have argued for the need to estimate farmers’ surpluses from agri-environment schemes as a first step towards improving policy evaluation.

In this paper, we apply propensity score matching and difference-in-difference (DID) analysis to estimate costs of farmers’ participation in the Entry Level Stewardship scheme. DID analysis allows us to estimate the total impact on farm business incomes associated with entry into an agri-environment scheme. The following section explains the problem of assessing the impact of policy on individual or firm behaviour and the way in which our methodology addresses it. In Section 'Take Up of ELS amongst Cereal Farms in Eastern England' we explain the context of the analysis and the data available, while in Section 'Examining ELS Impact on Income' we describe the methodology. Section 'Discussion and Conclusions' presents the results and Section 6 provides discussion and conclusions.

2. The Policy Evaluation Problem and Difference-In-Difference Approach with Propensity Score Matching

  1. Top of page
  2. Abstract
  3. 1. Agri-environmental Policy in England
  4. 2. The Policy Evaluation Problem and Difference-In-Difference Approach with Propensity Score Matching
  5. 3. Take Up of ELS amongst Cereal Farms in Eastern England
  6. 4. Examining ELS Impact on Income
  7. 5. Discussion and Conclusions
  8. References

Policy evaluation generally takes the form of a comparison of the outcomes of units treated by a policy intervention with some standard, a counterfactual, that may be determined by different methodologies. Among the various methodologies available, a difference-in-difference (DID) approach with propensity score matching (PSM) is effective in reducing time varying bias and selection bias that arise with simple comparisons, (e.g. before-and-after and cross-section approaches). The theoretical framework for DID is well discussed in the literature, notably by Imbens and Wooldridge (2009), Mueser et al. (2007), Smith and Todd (2005) and Dehejia and Wahba (2002) – which all refer back to Rosenbaum and Rubin (1983), who were the first to develop the PSM approach. This section outlines the biases arising from a simple comparison and explains the effectiveness of DID with PSM to address them in the context of our ELS evaluation, which draws on an administrative (and thus non-experimental) data source.

The expected overall policy impact can be determined as the average of the treatment effects on the treated units, in our context the difference in farm income (inline image) between the income of farms in ELS and the income that similar farms would have achieved had they not been in the ELS as:

  • display math(1)

where inline image and inline image are incomes of farm i in ELS (1) and not in ELS (0), respectively. N1 represents the number of ELS farms in the examination. The critical issue for the evaluation is that inline image cannot be observed for the ELS farms, i.e. the income that an ELS farm i would have generated had it not received the treatment and all other factors had remained the same. This unobservable counterfactual outcome needs to be substituted by some proxy that can be measured. A simple before-and-after comparison, for example, draws on the pre-programme position to impute a proxy for the counterfactual.

The impact estimator is inevitably influenced by time varying factors – for example, farms can be subject to cyclical business change or they could improve productivity over the observation period. Theoretically, the time variant bias might be addressed by eliminating the baseline income change that is unrelated to participation in ELS over the observation period, but the best observable baseline income change that we have available is that of a non-ELS farm as:

  • display math(2)

where the subscript t denotes the time of the treatment, the subscript s denotes a period of time after the treatment with  0. A farm j belongs to the control (i.e. non-ELS farm) group. The income change of the control farm is expected to project the ELS farm's income change component independent of the programme. The approach can be less biased either in an experimental setting (e.g. field trials with an experimental design so as to establish control plots that are subject to the same factors that potentially influence the characteristic of the unit that is of interest – such as, in the agricultural context, soils, weather or pests) or on the condition that treatment and control groups are randomly selected.3

However, our Farm Business Survey data are non-experimental, which may well include selection bias. Because ELS is voluntary, volunteers may well be different from the non-participants, exhibiting different cost structures or being differently motivated and/or capable. A farm's characteristics (e.g. being economically robust or managerially efficient enough to deal with the impact of ELS activities or large enough to allocate resources to ELS activities) in the pre-treatment period might affect the likelihood of entry to ELS and thus influence its subsequent income. There is thus likely to be adverse selection (Moxey, et al., 1999; Quillérou and Fraser, 2010) – farmers with lower compliance costs are more likely to apply than others.

To alleviate the selection bias, a treatment unit needs to be matched individually with control units that are as similar as possible in observable respects that are critical to programme participation and to the subsequent outcome. For example, in the evaluation of the US labour training programme, Mueser et al. (2007) matched the participants with the control labour force with respect to standard demographic characteristics such as age, education, race and labour market experience. In addition, they also maintained that where programme eligibility is limited, factors influencing eligibility can be a key aspect for matching. Analogously, our analysis employs farms’ physical and financial characteristics and farmers’ demographic attributes in matching. The DID method then assumes that any unobservable determining differences are constant over time, i.e. that differences at the start of the period of analysis remain the same throughout the period. Chabé-Ferret and Subervie (2013) provide a fuller discussion of the assumptions underpinning the analysis of agri-environment schemes using DID.

Multi dimensional matching is dealt with by a propensity score (PS), first introduced by Rosenbaum and Rubin (1983). A PS is the probability that any individual unit will be subject to the treatment and it is obtained by logit or probit regression drawing on the multiple characteristics of the units in the pre-treatment period as covariates. Observational units are then matched according to PS, although there are alternative approaches available in the selection of “closest” matches. The alternatives and their advantages and disadvantages are well summarised elsewhere, notably in Smith and Todd (2005) and Dehejia and Wahba (2002). However, in any approach, employing units only in the common support region is critical to alleviate biases. The common support region is an area where the PS distribution of the treatment group and that of the control group overlap. This restriction often achieves similar evaluation results between different matching methods.

With a series of methods for alleviating biases inherent in a simple comparison, our evaluation of ELS impacts on farm incomes drawing on observable information from the non-experimental dataset needs to translate the evaluation from (1) as:

  • display math(3)

where G1 and G0 represent the treatment group (i.e. ELS farms) and the control group (non-ELS farms), respectively. C1 denotes the subset of ELS farms in the common support region and C0 is the equivalent to the non-ELS farms. n1 is the number of ELS farms in the subset. Wi,j is the weight given to the jth unit in the control subset.

DID analyses with PSM have been applied in a variety of policy areas, and have a long history. Lechner (2010) refers to a study by Snow in 1854 where he compared the incidence of cholera in districts where the water supply was changed to water abstracted further upstream with the incidence in those districts where it was not changed in order to demonstrate that cholera was transmitted by water rather than by air. The first modern use of DID is usually attributed to Ashenfelter (1978) (Imbens and Wooldridge, 2009, p. 67) in estimating the impact of training programmes on earnings, and the approach has subsequently been extensively applied in the evaluation of training and labour market policy. The general approach to programme evaluation has also been used in relation to agricultural and environmental policies. Pufahl and Weiss (2009) used propensity score matching with DID analysis to assess the impact of agri-environment programmes in Germany, finding a positive effect on the area of agricultural use and a reduction in stocking densities and chemical use per hectare. Chabé-Ferret and Subervie (2012, 2013) have used DID analysis to estimate the causal effect of agri-environment programmes in France, finding that the windfall effects of the programmes depend on the specific requirements of the particular programme. Petrick and Zier (2011) apply a DID estimator to assess the employment effects of the full range of CAP measures in three East German states. They find that the investment aids and transfers to less-favoured areas had a zero marginal employment effect. Sneeringer (2011) has used a DID approach to examine the effects of environmental regulations on the structure of the Californian dairy industry, finding evidence that regulation has led to a decline in the size of the industry in Southern California where there are more restrictive regulations. Jarait≐ and Kažukauskas (2012) used a cross-country DID analysis and concluded that cross-compliance rules lead to a significant reduction in farm expenditure on fertiliser and pesticide. Sauer et al. (2012) have estimated distance frontier and propensity score matching models for cereal farms in the UK in investigating the effects of the Environmental Stewardship Scheme and Nitrate Vulnerable Zones on individual producer behaviour with respect to intensity, productivity and the structure of production. They conclude that farms enrolled in agri-environment schemes efficiently adjust their production decisions, becoming more diversified with regard to their production structure.

3. Take Up of ELS amongst Cereal Farms in Eastern England

  1. Top of page
  2. Abstract
  3. 1. Agri-environmental Policy in England
  4. 2. The Policy Evaluation Problem and Difference-In-Difference Approach with Propensity Score Matching
  5. 3. Take Up of ELS amongst Cereal Farms in Eastern England
  6. 4. Examining ELS Impact on Income
  7. 5. Discussion and Conclusions
  8. References

3.1. Farm Business Survey

Our evaluation draws on data for cereal farm businesses taken from the Farm Business Survey (FBS) for eastern England (the East Midlands, the East of England and the South East Government Office Regions) from 2004 to 2009.4 The FBS is a detailed national survey of farm businesses in England commissioned by the UK Department for Environment, Food and Rural Affairs (Defra), contributing to the European FADN data system. It is undertaken by Rural Business Research, a consortium of six British universities and colleges. The FBS collects data on a comprehensive range of financial and physical production measures from a panel sample of farm businesses in annual face-to-face interviews and accounts analysis. Cereals farms are defined as farms on which cereals, oilseeds, peas and beans harvested dry and land set aside account for over two thirds of their total Standard Gross Margin (a measure of farm structure) (Defra, 2010). From the cereal farms included in the FBS over the 6 year period, we exclude five with potato production for which discontinuities in returns caused exceptional changes in financial output. Fifty-two farms experiencing major changes in scale (greater than 5% or 2 or more hectares) over the period are also excluded, as well as those farm businesses for which a complete breakdown of enterprise performance was not collected. This ensured that we have effectively compared the same farm businesses through time. Five farms already in the more intensive Higher Level Stewardship agri-environment scheme are also excluded, since these farms are also likely to be outliers in other respects.

3.2. Development of participation in ELS

Thirty-four farms, 14.0% of the 242 cereal farms sampled in the FBS, entered ELS in 2005, the year in which it was introduced. Participation increased to 56.3% (130 farms) and 73.2% (161 farms) in the following 2 years, and has remained at around 75% subsequently. Table 1 presents a comparison between the ELS and non-ELS farm groups with respect to four income indicators – agricultural gross margin (AGM), agricultural output (ANM), farm business gross margin (FBGM) and farm business net income (FBNI). The first two measures relate more narrowly to agricultural output (AO) while the latter two cover all farm business output (FBO) including non-agricultural income. There are two versions of the income indicators for ELS farms – either including or excluding the ELS payment. However, the simple cross-sectional comparisons fail to identify an impact of ELS on farm income because of the unaddressed selection biases. ELS farms did not consistently underperform in terms of the income indicators in each year over the observation period. The annual income fluctuations within each group also indicate the possible biases inherent in a simple before-and-after comparison.

Table 1. Annual average income indicators for ELS and non-ELS farms
  ELS farmsNon-ELS farms
Income indicatorYearGrossNet of ELS payment  
  1. N for ELS farms and non-ELS farms: (208, 34), (130, 101), (161, 59), (176, 56), and (172, 55) for each year. Standard deviations are in parentheses.

  2. Source: Authors’ calculation based on the FBS.

AGM / AO (%)200543.83(23.46)40.55(23.77)42.24(30.63)
ANM / AO (%)2005–30.44(38.55)–33.71(39.44)–37.87(57.71)
FBGM / FBO (%)200565.43(11.08)63.45(11.02)64.97(14.40)
FBNI / FBO (%)20056.35(17.69)4.36(17.80)5.63(21.05)

4. Examining ELS Impact on Income

  1. Top of page
  2. Abstract
  3. 1. Agri-environmental Policy in England
  4. 2. The Policy Evaluation Problem and Difference-In-Difference Approach with Propensity Score Matching
  5. 3. Take Up of ELS amongst Cereal Farms in Eastern England
  6. 4. Examining ELS Impact on Income
  7. 5. Discussion and Conclusions
  8. References

4.1. Propensity scores by logit regression

The propensity score, which is the probability of farms participating in ELS, is obtained from a logit regression drawing on covariates that may affect the participation decision and influence subsequent farm income. The model is:

  • display math(4)

where pk,t denotes the probability of a farm k participating in the ELS scheme in period t; X is a matrix of farm characteristics in period t – 1; α is a coefficient matrix corresponding to X; and ε is an error term.

The ELS farms in our test must have FBS records available for three consecutive years, i.e. a pre-ELS year (t − 1), a year of joining ELS (t) and a post-ELS year (t + 1), and remain in the scheme in the third year. Since the FBS is neither a fully longitudinal survey nor a census, the number of such farms is somewhat limited with the result that 57 farms in the 2004 to 2008 FBS met this condition. Farms that entered ELS for the first time in 2009 could not be included in the analysis due to unavailability of information at t + 1. The sampled ELS farms’ pre-entry year (t − 1), therefore, ranged between 2004 and 2007. The number of the control farms with the 3-year records was 89.5 Our analysis assumes that the so-called Ashenfelter dip, the phenomenon that the outcome of the treatment unit falls shortly before the treatment (Ashenfelter, 1978), was not present at t − 1.6 Generally, it takes a couple of months from announcing the scheme to signing up so that a farm's performance is unlikely to be affected a year in advance in anticipation of entering ELS.

The covariates are area (total size and a proportion each of utilised agricultural area, enterprise area and owner-occupied area), unit labour input, farm business output components (agricultural output, non-farming miscellaneous income and income from agri-environmental activities (AES) other than ELS), independent income category, region dummy, business type dummy and year dummy.7 The reference farmer's attributes (age and education) are also included.8 Table 2 presents the measurement units and statistics of these variables.

Table 2. Descriptive statistics on characteristics of the ELS farms and non-ELS farms at t − 1
  ELS farmsNon-ELS farms
VariableDescription and unitMeanMinMaxMeanMinMax
  1. The standard deviations in parentheses. = 57 (ELS) and 89 (non-ELS). Reference levels for dummies are East Midland for Region, Partnership for Business type, Any level between School Only and Degree for Education, and t – 1 = 2004 for Year.

Total area(hectare; log form)5.283.927.114.903.436.83
(0.69)  (0.72)  
Utilised agri-area (UAA) ratioUAA / total area (%)95.1754.00100.0095.9670.1599.66
(6.59)  (4.51)  
Enterprise area ratioEnterprise area / total area (%)74.240.0098.1574.450.00100.00
(28.81)  (26.27)  
Owner-occupation ratioOO area / total area (%)65.670.00100.0074.520.00100.00
(38.62)  (36.55)  
Labour input ratioHours worked / total area (hours per hectare)21.420.2754.3424.451.4973.54
(11.52)  (12.48)  
Agriculture output (AO) ratioAO / farm business output (%)69.1339.92100.0070.694.98100.00
(17.21)  (20.31)  
AO ratio * year (2005) 39.720.0082.5216.470.0080.01
(30.88)  (27.53)  
AO ratio * year (2006) 3.330.0074.0715.150.0081.28
(14.43)  (29.11)  
AO ratio * year (2007) 2.160.0075.6614.320.0088.35
(11.73)  (29.87)  
Miscellaneous income (MI) ratioMI / FBO (%)40.560.8897.1931.690.6988.67
(20.47)  (18.50)  
MI ratio * year (2005) 30.730.0083.7413.150.0084.81
(25.82)  (22.40)  
MI ratio * year (2006) 2.240.0054.238.070.0088.67
(9.80)  (16.51)  
MI ratio * year (2007) 1.650.0053.935.560.0081.71
(8.84)  (13.47)  
AES payment ratioAES payment / FBO (%)2.790.0034.051.390.0011.09
(5.95)  (2.38)  
AES payt ratio * year (2005) 1.660.0023.690.320.006.58
(3.97)  (1.11)  
AES payt ratio * year (2006) 0.340.0010.
(1.79)  (1.20)  
AES payt ratio * year (2007)
(0.00)  (1.66)  
Independent incomedummy (1, if ≥ £5,000)0.07010.0901
(0.26)  (0.29)  
Region: South Eastdummy0.32010.3901
(0.47)  (0.49)  
East of Englanddummy0.37010.4301
(0.49)  (0.50)  
Business type: sole traderdummy0.40010.3301
(0.50)  (0.47)  
farming companydummy0.16010.1101
(0.37)  (0.32)  
Education: school onlydummy0.23010.4501
(0.42)  (0.50)  
degree or higherdummy0.21010.1801
(0.41)  (0.39)  
(11.78)  (10.96)  
Year: t − 1 = 2005dummy0.65010.2901
(0.48)  (0.46)  
t − 1 = 2006dummy0.05010.2301
(0.23)  (0.42)  
t − 1 = 2007dummy0.04010.2001
(0.19)  (0.40)  

Table 3 presents the results,9 which indicate that the farms more likely to participate in ELS were those with a relatively large enterprise, larger proportion of agricultural output and independent income. Sole traders and farming companies also tended to be in the scheme. On the other hand, farms in the South East or the East of England were less likely to enter ELS, compared with those in the East Midlands. Farms whose reference farmer was educated only at school also had a lower probability of joining.

Table 3. Logit model test result (dependent variable 1 = in ELS at t)
Covariate α Standard Error
  1. –2 Log likelihood ratio = 101.143. R2 = 0.475 (Cox & Snell) and 0.645 (Nagelkerke). Hosmer & Lemeshow χ2  = 2.891.

Total area1.04*0.58
UAA ratio–0.030.08
Enterprise area ratio0.04**0.02
Owner-occupation ratio0.000.01
Labour input ratio–0.050.03
Agriculture output (AO) ratio0.22***0.08
AO ratio * year (2005)–0.060.10
AO ratio * year (2006)–0.31**0.12
AO ratio * year (2007)12.43638.08
Miscellaneous income (MI) ratio0.13*0.07
MI ratio * year (2005)–0.010.09
MI ratio * year (2006)–0.140.10
MI ratio * year (2007)25.921,189.80
AES payment ratio0.050.14
AES payment ratio * year (2005)0.260.20
AES payment ratio * year (2006)0.60*0.31
AES payment ratio * year (2007)–69.293,347.23
Independent income2.16*1.16
Region: South East–3.04***1.03
East of England–2.28**1.12
Business type: sole trader1.61**0.66
farming company1.73*1.01
Education: school only–1.49**0.73
degree or higher–1.240.85
Year: (t − 1 = 2005)8.2710.12
(t − 1 = 2006)26.03**11.32
(t − 1 = 2007)–1,966.7793,581.93

4.2. Matching

We apply caliper matching where each ELS farm is matched with the closest non-ELS farm in terms of the PS. But to prevent closest yet poor matches, we employ only matched pairs whose PS difference is within a predetermined threshold, termed a caliper – in our case, 0.06.10 We allow replacement of non-ELS farms which helps reduce the PS distance. This is a one-to-one matching method, which means that a weight for an income change of a matched non-ELS farm, Wi,j in (5), is 1 and is 0 for those of all the remaining non-ELS farms. All the matches are undertaken in the common support region with the result that there are 39 pairs in our DID analysis. The matches were further checked by a balancing test – if the matching was implemented reasonably well, there is no significant difference between the selected ELS farm and non-ELS farm groups in terms of each of the means of the covariates employed in the logit regression to produce the PS. The test result represents a reasonably balanced outcome – there is no longer any significant difference between the treatment and control groups at a 5% level (Table 4).

Table 4. Balancing test result
 Mean DifferenceStd. Error of Differencet-value
  1. = 39. All farms with t − 1 = 2007 were not selected in the matching process. * 10% significance level.

Total area0.000.13–0.02
UAA ratio0.181.410.13
Enterprise area ratio–0.106.11–0.02
Owner-occupation ratio–3.287.82–0.42
Labour input ratio–0.452.40–0.19
Agriculture output (AO) ratio4.624.601.00
AO ratio * year (2005)5.166.990.74
AO ratio * year (2006)–4.503.80–1.18
AO ratio * year (2007)
Miscellaneous income (MI) ratio–6.104.62–1.32
MI ratio * year (2005)–0.955.62–0.17
MI ratio * year (2006)–7.144.36–1.64
MI ratio * year (2007)
AES payment ratio0.881.010.88
AES payment ratio * year (2005)0.77*0.401.91
AES payment ratio * year (2006)–0.320.37–0.86
AES payment ratio * year (2007)
Independent income dummy0.08*0.041.78
Region: South East0.050.110.49
East of England–0.030.12–0.22
Business type: sole trader0.050.110.48
farming company0.000.080.00
Education: school only0.130.101.32
degree or higher0.000.080.00
Year: (t − 1 = 2005)
(t − 1 = 2006)–0.100.07–1.50
(t − 1 = 2007)

Impact estimates are often highly sensitive to the estimator chosen (Smith and Todd, 2005, p. 306), so we examined changes in 16 income indicators – four core income indices (agricultural gross margin (AGM), agricultural net margin (ANM), farm business gross margin (FBGM) and farm business net income (FBNI)): each divided by total farm output (FO), total area (TA), utilised agricultural area (UAA) and labour input in terms of hours worked (HW). The total area and utilised agricultural area were measured in hectares while the labour input was defined as the total hours worked by the labour force on each farm. All the financial variables were in real terms, deflated by the Retail Price Index for cereals from the Office for National Statistics. Each of the income indicators for the ELS farms has two levels – with and without ELS payment. Table 5 presents these income indicators at t − 1, t and + 1. This shows that between t − 1 and + 1 all indicators increased in both groups, even when ELS payments were excluded for the treatment group indicators. The increases for the non-ELS group suggest a generally rising trend in farm incomes. Between t − 1 and t, the ELS farm group saw two income variables (AGM/HW and ANM/HW) decrease even when including ELS payment but five indicators (FBGM/FBO, AGM/TA, ANM/TA, AGM/UAA and ANM/UAA) decreased only when ELS payment was excluded. The control group did not see decreases in any indicator for the period.

Table 5. Average incomes before and after entering the ELS schemes for the ELS and non-ELS farm groups: Gross income on upper rows; income net of ELS subsidy on lower rows
Income indicator witht − 1t t + 1
measurement unitELSnon-ELSELSnon-ELSELSnon-ELS
  1. = 39 each for the treatment and control groups. The price variables are in real terms deflated by the retail price index of cereal.

AGM / agriculture output (AO) %44.0550.9145.6251.5554.9157.25
   41.71 51.09 
  (25.11) (20.01) 
ANM / AO %–23.72–36.95–18.55–21.874.574.73
   –22.46 0.75 
  (38.57) (35.89) 
FBGM / farm business output (FBO) %63.3668.0265.7672.0168.6372.06
   63.35 66.05 
  (12.79) (13.74) 
FBNI / FBO %6.869.7414.0319.5322.2521.08
   11.62 19.67 
  (13.58) (13.30) 
AGM / total area (TA)£ per hectare267.33280.46272.81316.25403.83451.31
   253.65 379.51 
  (166.38) (237.06) 
ANM / TA £ per hectare–68.17–110.56–56.28–60.6987.57110.91
   –75.44 63.25 
  (181.51) (281.99) 
FBGM / TA £ per hectare502.69582.62559.75739.67689.16817.78
   540.59 664.85 
  (169.40) (246.39) 
FBNI / TA £ per hectare60.87110.15127.21227.54234.49300.83
   108.05 210.18 
  (131.64) (177.43) 
AGM / utilised agri-area (UAA) £ per hectare283.30294.44287.26333.45424.20478.49
   266.85 398.41 
  (170.71) (242.45) 
ANM / UAA £ per hectare–74.98–116.05–65.43–63.0083.81122.58
   –85.84 58.02 
  (202.81) (307.28) 
FBGM / UAA £ per hectare531.60614.39591.97778.47727.17865.14
   571.56 701.38 
  (181.47) (257.56) 
FBNI / UAA £ per hectare64.28116.06134.05237.57244.25323.06
   113.63 218.45 
  (135.99) (182.68) 
AGM / hours worked (HW)£ per hour worked36.6212.8720.0214.4936.3220.85
   15.78 29.05 
  (124.38) (111.09) 
ANM / HW £ per hour worked–18.27–5.08–31.60–3.55–14.994.55
   –35.84 –22.26 
  (126.00) (95.62) 
FBGM / HW £ per hour worked99.3326.97108.2434.45122.0038.22
   104.00 114.72 
  (351.76) (340.88) 
FBNI / HW £ per hour worked8.664.7920.559.9333.4813.50
   16.31 26.21 
  (77.27) (69.76) 

4.3. Income change from t − 1 to t (i.e. year of entry into ELS)

4.3.1. Income including ELS payment

Firstly, we compare average income differences between the year of entry into ELS (t) and the year earlier (t − 1) for the treatment and control groups. The left half of Table 6 shows the results for the total income including ELS payment; with the second column for the average income difference for the treatment group and the third for the control group. The fourth column shows the DID, which is negative where the income of the treatment group increased less than that of the control group. Of 16 variables for income including ELS payment, the great majority (13) show a negative DID. Of those, four are statistically significant, indicating a negative impact of entry into ELS on these indicators, representing the two measures of total farm business income (FBGM and FBNI) per unit of total area and UAA. These figures suggest that FBGM on the ELS farms grew by the equivalent of £100 per hectare of TA and £104 per hectare of UAA less than that of non-ELS farms. In terms of FBNI, growth was £51 and £52 lower per TA and UAA, respectively. With respect to income solely from agricultural activities, all but one of the eight indicators show a negative DID, but are statistically insignificant. Similarly, there were no significant differences in the indicators of income per unit of labour input or in any of the other indicators.

Table 6. Test result: Mean differences between t − 1 & t for the treatment and control groups, and difference-in-difference between groups
Inc ELS paymentDELSDnon–ELSDIDNet of ELS paymentDELSDnon–ELSDID
  1. = 39 for each group. The standard error of DIDs can be found in parentheses. ***, ** and * indicate significance levels of 1%, 5% and 10%, respectively. The measurement units are percentage points for the variables divided by AO or FBO, £ per hectare for those divided by area and £ per hour for those divided by labour input.

AGM / AO1.570.650.92AGM / AO –2.340.65–2.98
  (4.59)  (4.64)
ANM / AO 5.1715.08–9.90ANM / AO 1.2615.08–13.81
  (9.52)  (9.59)
FBGM / FBO2.393.99–1.60FBGM / FBO –0.023.99–4.01***
  (1.42)  (1.45)
FBNI / FBO 7.179.79–2.62FBNI / FBO 4.769.79–5.03*
  (2.72)  (2.76)
AGM / TA5.4835.80–30.32AGM / TA –13.6835.80–49.48
  (34.59)  (34.43)
ANM / TA11.8949.86–37.97ANM / TA –7.2749.86–57.13
  (36.22)  (36.17)
FBGM / TA 57.06157.05–100.00***FBGM / TA 37.90157.05–119.16***
  (28.03)  (28.07)
FBNI / TA66.34117.39–51.05*FBNI / TA 47.18117.39–70.21**
  (28.78)  (28.89)
AGM / UAA3.9739.01–35.04AGM / UAA –16.4539.01–55.46
  (36.14)  (36.07)
ANM / UAA 9.5653.05–43.49ANM / UAA –10.8653.05–63.91*
  (37.95)  (38.00)
FBGM / UAA 60.38164.07–103.70***FBGM / UAA 39.96164.07–124.11***
  (28.65)  (28.72)
FBNI / UAA 69.77121.51–51.74*FBNI / UAA 49.35121.51–72.16**
  (29.35)  (29.48)
AGM / HW–16.601.61–18.21AGM / HW –20.841.61–22.45
  (30.40)  (31.42)
ANM / HW –13.331.53–14.86ANM / HW –17.571.53–19.10
  (30.74)  (31.57)
FBGM / HW 8.917.481.43FBGM / HW 4.677.48–2.81
  (12.12)  (12.35)
FBNI / HW11.895.146.75FBNI / HW7.655.142.51
  (16.08)  (15.93)
4.3.2. Income net of ELS payment

The right half of Table 6 shows results for income net of ELS payment. Again, most income indicators show a negative DID, with the exception of FBNI/HW. Amongst those with negative DIDs, seven indicators are significant, suggesting that ELS farms would have experienced lower incomes in the absence of the ELS payment. Of these, four indicators (FBGM/TA, FBNI/TA, FBGM/UAA and FBNI/UAA) were previously seen to be negative even taking account of the ELS payment. The remaining three (FBGM/FBO, FBNI/FBO and ANM/UAA) were negatively affected only after the deduction of the payment. The growth of FBGM/FBO and FBNI/FBO of the ELS farms were below those of the non-ELS farms by 4 and 5 percentage points, respectively, and the figures for FBGM/TA and FBNI/TA were lower by £119 and £70 per hectare, respectively. On average, FBGM/UAA, FBNI/UAA and ANM/UAA of the ELS farms grew less by £64, £124 and £72 per hectare than the equivalents for the non-ELS control group.

The findings suggest that the ELS activities were associated with the use of land resources particularly for overall farm business income. As in the case where the ELS payment was included, income per-unit labour use variables and those measured by AGM failed to show a significantly negative DID. This implies that the ELS activities did not require substantially more labour to be allocated for farm/agriculture business – farmers perhaps were engaged in the ELS activities in their own time without increasing reported labour use.

4.4. Income change from t to t + 1(after joining the ELS)

Table 7 shows the results for the period t to + 1, i.e. the year following entry into the ELS. Again, the left-hand columns relate to income including ELS payment and the right-hand columns to income net of ELS payment.

Table 7. Test result: Means difference between t and + 1 for the treatment and control groups, and difference-in-difference between groups
Inc ELS payment DELSDnon–ELSDIDNet of ELS paymentDELSDnon–ELSDID
  1. As for Table 6.

AGM / AO10.876.344.52AGM / AO 7.056.340.71
  (4.51)  (4.55)
ANM / AO 28.2941.68–13.39ANM / AO 24.4741.68–17.21
  (11.27)  (11.33)
FBGM / FBO5.274.051.23FBGM / FBO 2.694.05–1.36
  (2.04)  (2.05)
FBNI / FBO 15.3911.354.04FBNI / FBO 12.8111.351.46
  (4.55)  (4.57)
AGM / TA136.50170.86–34.36AGM / TA 112.18170.86–58.68
  (48.85)  (48.89)
ANM / TA155.74221.47–65.73ANM / TA 131.43221.47–90.04
  (56.47)  (56.57)
FBGM / TA 186.47235.16–48.69FBGM / TA 162.16235.16–73.00*
  (43.39)  (43.39)
FBNI / TA173.63190.68–17.05FBNI / TA 149.31190.68–41.37
  (46.45)  (46.51)
AGM / UAA140.91184.04–43.14AGM / UAA115.11184.04–68.93
  (51.74)  (51.83)
ANM / UAA 158.80238.63–79.83ANM / UAA 133.00238.63–105.63*
  (60.29)  (60.50)
FBGM / UAA 195.57250.75–55.18FBGM / UAA 169.78250.75–80.97*
  (45.96)  (45.97)
FBNI / UAA 179.97207.00–27.03FBNI / UAA 154.18207.00–52.83
  (49.17)  (49.29)
AGM / HW–0.307.98–8.27AGM / HW –7.577.98–15.54
  (25.20)  (27.05)
ANM / HW 3.289.63–6.35ANM / HW –3.999.63–13.62
  (27.02)  (28.27)
FBGM / HW 22.6711.2511.41FBGM / HW 15.3911.254.14
  (7.92)  (7.43)
FBNI / HW 24.828.7116.11FBNI / HW 17.558.718.84
  (14.06)  (12.58)
4.4.1. Income including ELS payment

With income including ELS, 11 income variables continue to show a negative DID, while five now show a positive DID. However, none of the DIDs were significantly different from zero, indicating that average income growth on the ELS farms was not significantly different from the control group a year after entering ELS. Thus, any possible negative impact was mitigated by the ELS payment. We may also note that there were no significantly higher incomes suggesting that the ELS payment is not significantly adding to farm incomes.

4.4.2. Income net of ELS payment

After taking out the ELS payment from the income figures, 11 indicators show a negative DID. Of these, three are statistically significant: FBGM/TA, ANM/UAA and FBGM/UAA, all of which also had a significantly negative DID in the previous comparison between t − 1 and t. This implies that without an ELS payment, the negative effect of ELS on these income measures would have been maintained for at least for 2 years after entering the scheme. However, the annualised DIDs for these variables, £36, £53 and £40 per hectare, respectively, are all lower than the equivalents observed in the comparison between t − 1 and t, suggesting that any negative impact of entering ELS diminishes over time. The income indicators based on FBNI no longer show statistical significance, which also implies a declining impact of the ELS. As in the analysis of the period t − 1 to t, there were no significant differences in the DIDs relating to indicators based on AGM or any expressed per unit of labour.

5. Discussion and Conclusions

  1. Top of page
  2. Abstract
  3. 1. Agri-environmental Policy in England
  4. 2. The Policy Evaluation Problem and Difference-In-Difference Approach with Propensity Score Matching
  5. 3. Take Up of ELS amongst Cereal Farms in Eastern England
  6. 4. Examining ELS Impact on Income
  7. 5. Discussion and Conclusions
  8. References

It is important to assess the impact of agri-environment schemes on farm incomes in practice. While the Farm Business Survey can identify the direct costs of undertaking environmental management, a significant proportion of the costs of entering a scheme relate to opportunity costs and transactions costs and so are not directly observable in the FBS. At the same time, simple comparisons between farms in and out of an agri-environment scheme suffer from the problem of selection bias. Difference-in-difference analysis addresses this problem. Matching pairs of cases, representing cases that have been treated and cases as controls, allows the estimation of counterfactual outcomes in a way that has generally not been possible in most types of social science research. The approach developed here has permitted an assessment of the full income effects of entering ELS on the farm business as a whole, including any direct, opportunity or transactions costs. The method is relatively data intensive and requires a comprehensive survey such as that provided by the FBS, including detailed information on the financial position of the farm business. However, even in this context there are still limitations in terms of data. The analysis has been undertaken with a relatively small sample of cereal farms that has prevented the analysis of years separately. We have allowed for this through the introduction of a year dummy variable to control for unobservables that may influence the likelihood of entering the scheme across the different years. We have similarly not been able to assess any potential differential impacts of the adoption of particular options within ELS. Notably, we assume that farms enter the ELS early within the financial year so that the costs of entering are incurred in the same year and that farms do not incur costs in the preceding year in preparation for entry in the following year. These are aspects that could be explored in future farm surveys. It would also be useful to undertake a similar analysis for other farm types.

Nevertheless we may draw a number of implications from the results. The evidence suggests that entry into ELS did have some negative impact on farm incomes, particularly on the overall farm business income representing both agricultural and non-agricultural sources. Given the insignificance of the impact on agricultural incomes, the greater impact appears to be on the non-agricultural element. Although we have no means of explaining this effect, it could reflect farmers focussing on the agri-environment scheme at the expense of efforts to generate other types of income on the farm, or perhaps that farmers choose to enter ELS in the first place as an approach to farm diversification, seeing it as offering an easier source of income than establishing a new non-agricultural enterprise. The results indicate that the losses per unit of agricultural land area were more often significant than per unit of labour. Thus the negative impacts were channelled through the utilisation of land resources rather than through increased use of labour. This might be expected in that it is clearly easier to make marginal adjustments in labour use than it is in the use of land. It has not been possible to analyse the individual ELS options separately, but this may reflect different costs of adopting the different options, several of which require changes in land use. Again this is something that could be explored further both for ELS and for other agri-environment schemes.

A comparison of the results for the two periods, the initial year of adoption and the year following adoption, suggests that the impact on farm incomes may diminish after an initial cost of implementing ELS. This may well reflect an initial cost of administration and adjustment to the requirements of the scheme, followed by lower costs, as adjustments are implemented. At this point, we do not have any information as to whether the costs might continue to fall beyond the two periods analysed here, but this should be assessed in future work. This raises the question of whether the level of payment offered might be higher in an initial period and then subsequently reduced. Of course, if the prospect of future benefits is the rationale for entry, then removing those future benefits could be likely to discourage farmers from entering the scheme. However, a voluntary agri-environment scheme that offers a flat rate payment based on average costs across heterogeneous land qualities and environmental values may lead to systematic over or underpayment (Fraser, 2009). The efficiency of the scheme depends on the relationship between costs and benefits within specific contexts. It might thus be suggested that a degree of targeting should be introduced into the ELS (Hodge and Reader, 2010) reflecting spatially differentiated social objectives for environmental improvement. The results here suggest less potential advantage from the introduction of competitive tendering, given the rather modest surpluses that entrants appear to enjoy. But of course, this may reflect the particular architecture of ELS and not apply to agri-environment schemes in general.

The results indicate that the ELS payment compensated for much of the cost of joining the scheme, although not sufficiently to fully compensate for costs in the first year. On the other hand, surprisingly given the ELS case studies discussed earlier and a general assumption that ELS offers considerable opportunity for adverse selection, we observe no evidence that entry into ELS significantly increased levels of farm income. Rather, the evidence points towards some modest level of cost, at least initially. Assuming that the methodology is sufficiently robust, we may thus reflect on why farmers might choose to enter a scheme that offers little financial benefit. It is of course possible that farmers are unaware of the full costs involved, or that they see an initial cost of entry as an investment generating returns in the longer term. But it is also possible that farmers are less driven by financial returns, and see personal benefits from the improvement to the environment on their farm – or the availability of the ELS motivates some sense of stewardship (Colman, 1994). One reason why farmers may join ELS is that they are required to so if they want to join HLS. Because HLS may be thought to lead to greater payments and profits, a small loss for ELS-related activities may be offset by the profit to be made from HLS in the future.

There is evidence that farmers’ environmental attitudes do influence farmers’ decisions to enter agri-environment schemes (e.g. Vanslembrouck et al., 2002; Defrancesco et al., 2008) or that the availability of schemes may itself influence farmers’ attitudes towards the environment (Wilson and Hart, 2001), in which case the voluntary approach may be more cost-effective than implied by an analysis that emphasises the problem of adverse selection. This echoes Quillérou et al.'s (2011) possible discovery of “auspicious” rather than adverse selection in the Higher Level Stewardship scheme. This points towards a possible inconsistency in the economic analysis of agri-environment policies that finds non-financial motivations for entry into schemes and yet tends nevertheless to assume that schemes “over-reward all but the marginal producer” (Hanley et al., 2012, p. 99). Some analyses have addressed this through multi-utility models, giving empirical support for approaches that recognise self interest combined with some wider social concerns reflecting stewardship or empathy with others (e.g. Chouinard, et al., 2008; Sheeder and Lynne, 2011). This suggests a promising way forward.

However, none of this offers much comfort to be drawn from the current CAP reforms (European Commission, 2013). The new environmental initiatives involve the introduction of standardised conditions for Pillar 1 payments, with potential reductions in funding available for more targeted agri-environment schemes under Pillar 2 (cf. Hodge, 2013a). Thus 30% of the direct support paid to farmers will be linked to requirements for “efficient agricultural practices aimed at preserving biodiversity, soil quality and the environment in general”.11 Also “restoring, preserving and enhancing ecosystems” is only one priority out of six in Pillar 2, and each country will draw up its own “national rural development strategy”, in light of its own situation. This indicates less flexibility and opportunity to target agri-environment expenditures towards changes in farm practices and environmental benefits that generate the greatest social benefit and hence less efficient agri-environment policies.

  1. 2

    Details of ELS are available at: (last accessed on 3rd July 2013).

  2. 3

    Dehejia and Wahba (2002), however, caution that an experimental setting could still be subject to bias, because of problems such as self-selection or some systematic judgment by the researcher in selecting units to be assigned to the treatment. Mueser et al. (2007) hold that the average programme effect for individuals subject to random assignment may be estimated as the simple difference in outcomes for those assigned to treatment and those assigned to the control group.

  3. 4

    One year is from 1st of April to the end of March in the following year.

  4. 5

    The control farms with four or more records were employed multiple times.

  5. 6

    The assumption is extended to reasonably parallel movements of the outcomes of the ELS farm group to the equivalents of the matched group without any irregular upward shift of the ELS farms at t or t+s to accommodate the dip.

  6. 7

    Independent income was reported as an ordered categorical variable in the data source.

  7. 8

    We tested characteristics available from the FBS. Of those, the above 28 variables explained the relationship well. Other covariates available from the FBS were examined but withdrawn from the model due either to lack of impact, possible collinearity, bias towards one group or failure at the final checking of the validity of matching.

  8. 9

    A number of studies have analysed the determinants of entry into agri-environment schemes. See for instance Hynes and Garvey (2009) or Quillérou et al. (2011).

  9. 10

    We set the caliper with reference to the standard deviation of the propensity scores of the examined farms in the common support region, which was 0.27.

  10. 11


  1. Top of page
  2. Abstract
  3. 1. Agri-environmental Policy in England
  4. 2. The Policy Evaluation Problem and Difference-In-Difference Approach with Propensity Score Matching
  5. 3. Take Up of ELS amongst Cereal Farms in Eastern England
  6. 4. Examining ELS Impact on Income
  7. 5. Discussion and Conclusions
  8. References
  • Ashenfelter, O.Estimating the effect of training programs on earnings’, Review of Economics and Statistics, Vol. 6, (1978) pp. 4757.
  • Baldock, D. and Lowe, P.The development of European agri-environment policy’, in: M. Whitby (ed.), The European Environment and CAP Reform. Wallingford: CAB International, 1996.
  • Chabé-Ferret, S. and Subervie, J.Econometric methods for estimating the additional effects of agri-environment schemes on farmers’ practices’, in OECD (ed). Evaluation of Agri-Environmental Policies: Selected Methodological Issues and Case Studies (Paris: Organisation for Economic Cooperation and Development, 2012).
  • Chabé-Ferret, S. and Subervie, J.How much green for the buck? Estimating additional and windfall effects of French agro-environmental schemes by DID-matching’, Journal of Environmental Economics and Management, Vol. 65, (2013) pp. 1227.
  • Chouinard, H., Paterson, T., Wandschneider, P. and Ohler, A.Will farmers trade profits for stewardship? Heterogeneous motivations for farm practice selection’, Land Economics, Vol. 84, (2008) pp. 6682.
  • Colman, D.Ethics and externalities: Agricultural stewardship and other behaviour’, Journal of Agricultural Economics, Vol. 45, (1994) pp. 299311.
  • Defra. The Rural Development Programme for England 2007–2013 (London: Department for Environment, Food and Rural Affairs, 2007).
  • Defra. Definitions of Terms used in Farm Business Management. (London: Department for Environment, Food and Rural Affairs, 2010).
  • Defrancesco, E., Gatto, P., Runge, F. and Trestini, S.Factors affecting farmers’ participation in agri-environmental measures: a Northern Italian perspective’, Journal of Agricultural Economics, Vol. 59, (2008) pp. 114131.
  • Dehejia, R. H. and Wahba, S.Propensity score-matching methods for non-experimental casual studies’, The Review of Economics and Statistics, Vol. 84, (2002) pp. 151161.
  • European Commission. Agri-environment measures. Overview on general principles, types of measures and application. Directorate General for Agriculture and Rural Development, Unit G-4 Evaluation of Measures applied to Agriculture, Studies (Brussels: European Commission, 2005).
  • European Commission. The Common Agricultural Policy (CAP) and Agriculture in Europe. MEMO/13/631 (Brussels: European Commission, 2013).
  • Falconer, K.Farm-level constraints on agri-environmental scheme participation: a transactional perspective’, Journal of Rural Studies, Vol. 16, (2000) pp. 379394.
  • Fraser, R.Land heterogeneity, agricultural income forgone and environmental benefit: an assessment of incentive compatibility problems in Environmental Stewardship Schemes’, Journal of Agricultural Economics, Vol. 60, (2009) pp. 190201.
  • Grey, P. and Jones, J. ‘The financial implications for farmers of Entry Level stewardship participation: Case based assessment in Gloucestershire’, in The 82nd Agricultural Economics Society Conference. Royal Agricultural College: Cirencester, 2008.
  • Hanley, N., Banerjee, S., Lennox, G. and Armsworth, P.How should we incentivize private landowners to ‘produce’ more biodiversity?’, Oxford Review of Economic Policy, Vol. 28, (2012) pp. 93113.
  • Harrison, G. and Jones, J.A farm-level assessment of the profitability of Entry Level Scheme participation in the Lincolnshire Wolds’, in The 84th Agricultural Economics Society Conference (University of Edinburgh, 2010).
  • Hodge, I. and Reader, M.The introduction of Entry Level Stewardship in England: Extension or dilution in agri-environment policy?’, Land Use Policy, Vol. 27, (2010) pp. 270282.
  • Hodge, I.Agri-environment policy in an era of lower government expenditure: CAP reform and conservation payments’, Journal of Environmental Planning and Management, Vol. 56, (2013a) pp. 254270.
  • Hodge, I.European agri-environmental policy: the conservation and re-creation of cultural landscapes’, in Duke , J. M. , Wu and J (eds). The Handbook of Land Economics (New York: Oxford University Press, 2013b forthcoming).
  • Hynes, S. and Garvey, E.Modelling farmers’ participation in an agri-environmental scheme using panel data: an application to the rural environment protection scheme in Ireland’, Journal of Agricultural Economics, Vol. 60, (2009) pp. 546562.
  • Imbens, G. and Wooldridge, J.Recent developments in the econometrics of program evaluation’, Journal of Economic Literature, Vol. 47, (2009) pp. 586.
  • Jarait≐, J. and Kažukauskas, A.The effect of mandatory agro-environmental policy on farm fertiliser and pesticide expenditure’, Journal of Agricultural Economics, Vol. 63, (2012) pp. 656676.
  • Jones, J.Assessment of the costs and benefits of agri-environment scheme participation pre- and post-decoupling’, in The 79th Agricultural Economics Society Conference. (University of Nottingham, 2005).
  • Langton, S. Cereals Farm: Economic Performance and Links with Environmental Performance, Department for Environment, Food and Rural Affairs Agricultural Change and Environment Observatory Research Report No. 25 (Defra, London, 2011).
  • Lechner, M.The estimation of causal effects by difference in difference methods’, Foundations and Trends in Econometrics, Vol. 4, (2010) pp. 165224.
  • Moxey, A., White, B. and Ozanne, A.Efficient contract design for agri-environment policy’, Journal of Agricultural Economics, Vol. 50, (1999) pp. 187202.
  • Mueser, P. R., Troske, K. R. and Gorislavsky, A.Using state administrative data to measure program performance’, The Review of Economics and Statistics, Vol. 89, (2007) pp. 761783.
  • Petrick, M. and Zier, P.Regional employment impacts of Common Agricultural Policy measures in Eastern Germany: a difference-in-difference approach’, Agricultural Economics, Vol. 42, (2011) pp. 183193.
  • Pufahl, A. and Weiss, C.Evaluating the effects of farm programmes: Results from propensity score matching’, European Review of Agricultural Economics, Vol. 36, (2009) pp. 79101.
  • Purvis, G., Louwagie, G., Northey, G., Mortimer, S., Park, J., Mauchline, A., Finn, J., Primdahl, J., Vejre, H., Vesterager, J. P., Knickel, K., Kasperczyk, N., Balazs, K., Vlahos, G., Christopoulos, S. and Peltola, J.Conceptual development of a harmonised method for tracking change and evaluating policy in the agri-environment: The agri-environmental footprint index’, Environmental Science and Policy, Vol. 12, (2009) pp. 321337.
  • Quillérou, E. and Fraser, R.Adverse selection in the Environmental Stewardship Scheme: Does the Higher Level Stewardship Scheme design reduce adverse selection?’, Journal of Agricultural Economics, Vol. 61, (2010) pp. 369380.
  • Quillérou, E., Fraser, R. and Fraser, I.Farmer compensation and its consequences for environmental benefit provision in the Higher Level Stewardship Scheme’, Journal of Agricultural Economics, Vol. 62, (2011) pp. 330339.
  • Rosenbaum, P. and Rubin, D.The central role of the propensity score in observational studies for causal effects’, Biometrika, Vol. 70, (1983) pp. 4155.
  • Sauer, J., Walsh, J. and Zilberman, D.Behavioural change through agri-environmental policies? A distance function based matching approach’, in The 86th Agricultural Economics Society Conference (University of Warwick, 2012).
  • Sheeder, R. and Lynne, G.Empathy-conditioned conservation: “Walking in the shoes of others” as a conservation farmer’, Land Economics, Vol. 87, (2011) pp. 433452.
  • Smith, J. A. and Todd, P. E.Does matching overcome LaLonde's critique of nonexperimental estimators?’, Journal of Econometrics, Vol. 125, (2005) pp. 305353.
  • Sneeringer, S.Effects of environmental regulation and urban encroachment on California's dairy structure’, Journal of Agricultural and Resource Economics, Vol. 36, (2011) pp. 590614.
  • Vanslembrouck, I., Van Huylenbroeck, G. and Verbeke, W.Determinants of the willingness of Belgian farmers to participate in agri-environmental measures’, Journal of Agricultural Economics, Vol. 53, (2002) pp. 489511.
  • Wallis, J. and Jones, J.The financial impact of the Entry Level Scheme in the uplands. A case study assessment on farms in Teesdale’, paper presented at the ROOTS Rural Research Conference, RICS, London, 2007.
  • Wilson, G. and Hart, K.Farmer participation in agri-environment schemes: Towards conservation-oriented thinking?’, Sociologia Ruralis, Vol. 41, (2001) pp. 254274.