We use historical spatial data on commons and land use outcomes sourced from the Biological Survey of the common lands in England (Defra 2012) and from the annual Farm Business Survey (Ministry of Agriculture, Fisheries and Food 2010) to investigate whether management by local commons associations is systematically associated with more sustainable grazing outcomes. We use two-stage least squares regression analysis to examine if commons associations have an effect on more sustainable grazing outcomes using economic heterogeneity as our instrumental variable.
Data Sources
The Biological Survey of the Common Lands in England
The biological survey of the common lands in England was compiled over a decade (1982–1993) by the Rural Surveys Research Unit at Aberystwyth University but only recently made publicly available. The survey contains extremely valuable biological, geographical, and regulatory data for 7052 parcels of common land sourced from the registers of common land, archives and field surveys.Footnote 10 716 parcels of common land in the biological survey include field survey evaluations on grazing intensity based primarily on vegetation condition and quantified according to a six-category classification. A score of 1 or 2 was given to parcels of common land with low grazing intensity. A score of 3 or 4 was given to parcels of common land with medium grazing intensity. A score of 5–6 was given to parcels with high grazing intensity. According to Aitchison et al. (2000), commons judged as having low-grazing intensity can be considered to be under-grazed and commons judged to have high intensity can be considered overgrazed. We transform this measure of grazing intensity into a binary variable with 0 representing either under- or over-grazed intensities and 1 representing sustainable grazing outcomes. Our rationale for this comes from the poor environmental outcomes associated with both under-grazing and over-grazing. For example, under-grazing tends to lead to species invasion and a loss of biodiversity. Similarly, over-grazing tends to reduce vegetation, expose underlying soil and cause soil erosion (Williams 2006).
Whether parcels were governed by local commons associationsFootnote 11 were recorded for all 716 parcels of common land with a total of 48 parcels being recorded as governed by voluntary local commons associations during the survey period. Additional variables of interest in the survey include: whether or not there were disputes over rights of common among commoners; the number of grazing rights per hectare; the number of different types of rightsFootnote 12 held; elevation; dominant habitat and livestock-type; and whether or not the parcel of common land falls partially or fully within the boundaries of an Area of Outstanding Natural Beauty (AONB), a Site of Special Scientific Interest (SSSI) or a National Park.
The Farm Business Survey
The farm business survey (FBS) is an annual survey that collects data from approximately 2300 farms in England and Wales. The survey covers the physical, environmental and economic performance of farm businesses (Defra 2017). We use spatial data from the farm business survey over the period 1982–1993 to construct measures of user attributes for our econometric analysis. Specifically, we construct measures of farmer age, farmer education, farm size, farm income, and environmental payments received (see Table 8). We also calculated our instrumental variable, non-livestock income heterogeneity of farms (economic heterogeneity), using data from the Farm Business Survey. To construct our measures, we first categorized individual farms by the lowest local unit of local administration in England (LAU1) for the years 1982–1993. For instance, taking our instrumental variable (economic heterogeneity) as an illustrative example, we calculated the Gini coefficient—a widely accepted measure of economic inequality—of non-livestock farm income for each LAU1 polygon over the period 1982–1993. We then averaged Gini coefficients for each LAU1 polygon over the period 1982–1993 before overlaying all parcels of common land. We assigned the Gini coefficient for each LAU1 polygon to the parcels of common land spatially located within respective LAU1 boundaries.Footnote 13
Econometric Estimation
We begin by estimating the relationship between the management of local commons and sustainable grazing outcomes using a linear probability model. While a basic linear probability model is instructive for determining whether a statistical relationship exists between voluntary commons associations and sustainable grazing outcomes it has its limitations. Most importantly, there is likely to be endogeneity between the existence of commons associations and environmental outcomes. That is, it may be that better organized and more community spirited commons may be more likely to form voluntary commons associations. In such a situation, it becomes difficult to determine whether it is the voluntary commons associations themselves producing better environmental outcomes or some other underlying characteristics of the common. In such a case, the linear probability estimation will produce biased estimates because the variable capturing whether the common is managed by a voluntary commons association is likely to be correlated with the error term of the regression. In response to the potential problem of endogeneity, we employ a two stage least squares model to estimate a linear relationship between sustainable grazing outcomes and voluntary commons associations.
Linear Probability Regression Analysis
We estimate the relationship between the management of local commons and sustainable grazing outcomes using a linear probability model that takes the following form:
$$ g_{i} = \alpha_{0} + \beta_{1} C_{i} + X_{i}^{\prime } {\Upphi } + \varepsilon_{i} $$
(1)
where gi is a binary measure for sustainable grazing outcomes for individual common i, Ci measures whether individual common i has a voluntary commons association in place for the period of measurement, and X
′
i
is a vector of control variables.Footnote 14 We choose a linear probability model to produce results that are more consistent with the two stage least squares estimation below. In selecting our control variables, we draw on Ostrom’s (2009) socio-ecological systems (SES) framework that identifies systems and sub-systems that affect the likelihood of resource users self-organising to achieve sustainable outcomes. Ostrom’s SES framework consists of core sub-systems covering users, governance systems and resource systems and units which interact to produce a range of sustainability outcomes. These core sub-systems are further disaggregated into second-level variables.
We follow Ostrom’s SES framework in grouping our control variables according to sub-system.Footnote 15 Starting with user-attributes, we include the number of users, farmer age, education, income and farm size. For governance systems, we include whether or not there were disputes over rights of common among commoners, the number of grazing rights per hectare, the number of different types of rights held, whether or not the parcel of common land falls partially or fully within the boundaries of an Area of Outstanding Natural Beauty (AONB), a Site of Special Scientific Interest (SSSI) or a National Park and average payments from the environmental sensitive areas (ESA) scheme. For resource systems and units, we include elevation, dominant habitat type, and livestock type.
Two-Stages Least Squares Regression Analysis
We employ a two stage least squares model to estimate a linear relationship between sustainable grazing outcomes and voluntary commons associations. The first and second-stage regressions are of the following form:
$$ C_{i} = \alpha_{0} + \gamma_{1} Z_{i} + X_{i}^{\prime }\Phi + u_{i} $$
(2)
$$ g_{i} = \alpha_{0} + \beta_{1} C_{i} + X_{i}^{\prime }\Phi + v_{i} $$
(3)
In these two equations, gi is the measure for sustainable grazing outcomes, X
′
i
is a set of exogenous control variables (as outlined above), Ci is the endogenous measure for whether a voluntary commons association in place, Zi is the instrumental variable (economic heterogeneity, which is discussed below), and ui and vi are the econometric errors, which contain unobservable factors that can either be related to the formation of voluntary commons associations, sustainable grazing outcomes, or both.
The parameter of interest is the second stage parameter \( \beta_{1} \), which aims to measure the causal effect of commons associations on sustainable grazing outcomes. For this parameter to be identified, the instrument of economic heterogeneity must be both ‘relevant’ and ‘valid’. Specifically, our instrument must be both correlated with the establishment of commons associations and have no effect on sustainable grazing outcomes other than through the establishment of commons associations.
Starting with the ‘relevance’ requirement, theoretical and empirical evidence suggests that we can expect economic heterogeneity to be associated with the establishment of local commons associations. While it is acknowledged that there is debate over the role that economic heterogeneity plays in relation to the promotion of community organizations, there seems little doubt that it is an important factor (e.g. Johnson and Libecap 1982; Tang 1994; Ruttan and Borgerhoff Mulder 1999; Vedeld 2000; Varughese and Ostrom 2001; Dayton-Johnson and Bardhan 2002; Kurien and Dietz 2004; Ruttan 2008; Andersson and Agrawal 2011; Ito 2012). In the context of the English commons we posit that economic heterogeneity allows for some ‘privileged’ commoners to be able to more easily absorb the start-up and running costs associated with a local commons association, and as will be seen in our results (Table 3), our measure of economic heterogeneity is strongly positively correlated with local commons associations.Footnote 16
Turning to the second requirement of a good instrument, ‘validity’. This requires that economic heterogeneity should exclusively affect sustainable grazing outcomes through its first stage impact on the establishment of commons associations. We consider this to be a defendable assumption in our case, as it is not apparent how economic heterogeneity (as we measure it—see below) could play a role in sustainable grazing outcomes, other than through better community organization that is manifested by the establishment of a commons association. While it is true that income levels may affect environmental preferences or capacity to undertake investments that protect the environment, there is no apparent reason why economic heterogeneity should. Given that economic heterogeneity may be linked to concentration of common use rights that may in turn be linked to grazing outcomes, we construct a measure that de-links this potential relationship.Footnote 17 We do this by excluding income derived from livestock in our measure of farm income that is used to construct our measure of economic heterogeneity. By excluding all income generated from livestock (whether on private or common land) we hope to exclude the potential for a direct relationship between income heterogeneity and management of common land, thus providing us with an instrument that should be valid.Footnote 18 As an additional check, we estimate an overidentified model using generalized method of moments (gmm) estimation. To do this, we disaggregate our economic heterogeneity variable into two components: crop income and off-farm income. We calculate the gini coefficient for each component and include the two measures of economic heterogeneity as our instruments. This allows for us to test the validity of our instruments using the Hansen J-statistic test.