The European Commission partially justifies the Common Agricultural Policy (CAP) of the EU with the CAP’s contribution to ‘viable rural communities’. Maintaining viable rural communities was one of the three strategic aims for the CAP set out in the Commission’s Communication on the CAP towards 2020 in November 2010. It was justified in the following terms:
To maintain viable rural communities, for whom farming is an important economic activity creating local employment; this delivers multiple economic, social, environmental and territorial benefits….
Agriculture is an integral part of the European economy and society. In terms of indirect effects, any significant cut back in European farming activity would in turn generate losses in GDP and jobs in linked economic sectors – notably within the agri-food supply chain, which relies on the EU primary agricultural sector for high quality, competitive and reliable raw material inputs, as well as in non-food sectors. Rural activities, from tourism, transport, to local and public services would also be affected. Depopulation in rural areas would probably accelerate. There would therefore be important environmental and social consequences.
This claim raises at least three important research questions. Does agricultural support actually promote agricultural employment? Does increased agricultural output and employment actually contribute to increases in non-agricultural output and employment? And might emphasising non-agricultural development be a more effective way of stabilising and enhancing the viability of rural areas?
The last two questions are examined in a recent report by Anne Margarian from the Johann Heinrich von Thünen-Institut in Germany and her findings are summarised in this post.
At first sight, assessing the relationship between changes in agricultural and non-agricultural activity seems rather straightforward. Why not simply examine the relationship between the two variables and draw the appropriate conclusion? If we do this, as shown in the Margarian paper, it turns out there is no obvious, linear relationship between the general development of total employment or GVA and the development of agricultural employment or GVA across regions. Positive correlations dominate the observed gross relationship between agricultural and non-agricultural changes in GVA and negative correlations dominate the relationship in the developments of employment.
However, such crude comparisons are flawed for a number of reasons:
The problem of spurious correlation. Common trends may be due to the simultaneous influence of a third factor rather than demonstrating causality. For example, the positive estimated coefficient of the GVA relationship mainly reflects the existing coincidence between a generally positive development of agricultural GVA on the one hand and of non-agricultural GVA on the other hand. This coincidence is not due to a causal relationship but rather to the generally rising technical efficiency of production. A similar non-causal coincidence underlies the negative estimated coefficients in the employment model, as in structural change agricultural employment generally declines, while non-agricultural employment more often grows. A more sophisticated analysis needs to be undertaken to uncover the causal relationships.
The gross effect is made up of more complex dynamic effects which may cancel each other out. Detailed analysis needs to differentiate different effects in order to interpret the observed gross relation between agricultural development and the development of other sectors. On the one hand, changes in agricultural employment or GVA may influence non-agriculture employment or GVA; but there may also be reverse influences from non-agriculture employment or GVA to agricultural employment or GVA.
On the other hand, while the Commission quotation above emphasises the complementary relationship between developments in agriculture and non-agriculture in any given region, the two activities can also be competitive. There is thus an ambiguous relationship between agricultural development and non-agricultural development due to the parallel existence of multiplier and income effects on the one hand and competition effects for scarce resources, i.e., land, labour and capital, on the other hand.
Margarian describes how this gives rise to a 2 x 2 matrix of possible effects:
Transmission effects
The induction/dependence effect (positive agriculture => non-agriculture). The induction effect is where growth in agriculture positively affects non-agriculture largely through the respending of additional income within the region. Where the same complementary relationship exists when agriculture is contracting (and thus pulling down activity in the non-agriculture sector), it is referred to as a dependence effect.
The stabilisation/destabilisation effect exists when changes in agriculture are complementary to changes in non-agriculture (positive non-agriculture => agriculture). If agriculture relies on part-time farms, for example, a positive development of incomes in the non-agricultural sector could stabilise these farms. Conversely, a crisis outside of agriculture could also threaten the viability of agriculture.
Competition effects
The mobility/immobility effect where changes in agriculture and non-agriculture are inversely related (negative agriculture => non-agriculture). Due to the competition for scarce resources, a positive [negative] development in agriculture, that implies immobility [mobility] of agricultural factors, induces a negative [positive] development of non-agricultural sectors, which otherwise benefit from freed factors from agriculture.
The attraction/detraction effect exists where there is an inverse relationship between changes in non-agriculture and agriculture (negative non-agriculture => agriculture), i.e. the attraction [detraction] of agricultural factors by a positive [negative] development in non-agricultural sectors.
Taking account of regional heterogeneity. These dynamic relationships are likely to be different and to be weighted differently in different types of regions. A statistical analysis needs to take account of the way these relationships change with respect to different regional characteristics. In this work, regions were classified based on quintiles from the indicators ‘agricultural productivity’ and ‘share of agricultural employees’. Combining these five x five classes created a finer classification of 25 classes. For example, of the total of 1,296 NUTS3 regions, Romania has 41 regions with both very high shares of agricultural employment but very low productivity, and 1 region with a very low share of agricultural employment and very low productivity. Thirty per cent of Hungary’s regions fall into the third quintile for agricultural employment share and the second quintile ranked by agricultural productivity, and so on. These regional classes were used to help unravel the relationships in the employment model; in the GVA model it was possible to introduce the underlying characteristics (share of agricultural employment and agricultural productivity) directly.
Finally, unravelling the underlying relationships is made particularly difficult in an EU context because of the inadequacy of data. Socio-economic data availability distinguishing between agriculture and non-agriculture across the NUTS3 regions is essentially limited to employment and GVA data, and only for a limited time period (2002-2008).
Methodology
These characteristics of the agriculture – non-agriculture relationship pose evident challenges for empirical research. The Margarian paper addresses these challenges in a particularly innovative way. Two separate models are estimated, one for the employment relationship and one for the GVA relationship. The variables employed in the two models are constructed from just four indicators: agricultural employment, non-agricultural employment, GVA of agriculture and non-agricultural GVA.
The variables are introduced in different ways and with varying interactions. In the agricultural employment model, for example, the dependent variable to be explained is the (scaled) change in non-agricultural employment. The explanatory variable is the change in agricultural employment, and this variable is introduced three times as a lagged, contemporaneous and leading variable. Quadratic terms are also introduced to capture non-linear relationships, and each of the three forms of the agricultural employment variable is interacted with the regional classification to capture the expected regional heterogeneity in the relationship between changes in agricultural and non-agricultural employment.
The introduction of agricultural employment as a lagged, contemporaneous and leading variable both controls for spurious correlation and allows for the identification of the different kinds of relationship between changes in agriculture and in other sectors. The effect of the lagged variable is interpreted as the effect of agricultural development on non-agricultural development, and the effect of the leading variable after appropriate transformation is interpreted as the effect of changes in the non-agricultural sector on agricultural development.
The current or contemporaneous effect is more difficult to interpret because it incorporates all of the identified effects and additionally the spurious relation between agricultural and non-agricultural developments within a region, i.e. those relations that are due to unobserved regional influences, which affect both developments simultaneously.
Findings
Findings from the employment model can be used to illustrate some outcomes from the analysis.
The estimated relation between the leading variable (t+1) of the agricultural change and current (t+0) non-agricultural change (lead effect) is negative in many classes. The author argues this result supports the interpretation of the lead effect as a pull effect that captures the attraction of agricultural factors by positive developments outside of agriculture.
The effect of the lagged change in agricultural employment upon non-agricultural employment (lag effect) is, in contrast, rather positive in many classes. This supports the existence of induction or dependence effects. A positive lag effect is mainly observed where agricultural productivity is low or the share of agricultural employment is high. A negative lag effect is of high relevance for regions with a very low share of agricultural employment and a very high agricultural productivity. Here the competition for scarce resources leads to the effect that restricted mobility of agricultural factors restricts non-agricultural development.
Significant negative contemporaneous relations between the agricultural and non-agricultural development (current effect) imply that the competition effects dominate the short-term relation between agriculture and other sectors in many regions.
The overall effect (what the paper calls the ‘all-up’ effect), which is the sum of the lead, the current and the lag effects is strongly negative. The author sees this as confirmation of the domination of the competition effect. However, including the current effect leaves this conclusion open to the objection that it merely reflects the likelihood of third factor effects and thus is a spurious causation. The independent evidence from the lag and lead variables probably deserves more weight. Even here, however, competition effects seem more important than transmission effects for many EU regions.
Different development regimes
The regionally differing relative strength of the competition effect as compared to the transmission effect is used to identify different regimes of developments of agriculture in specific regional environments.
Regions in France, Scandinavia and northern Italy are characterised by positive transmission effects in employment and negative competition effects in GVA.
Regions in Germany and Ireland, for example, are characterised by negative competition effects in employment and a positive induction effect in GVA.
Many eastern European regions are ascribed a negative attraction and a positive induction effect in employment and positive transmission effects in GVA.
These are highly simplified findings from a very complex analysis, and it is hard to know what policy conclusions to draw. Even in those regions where increased agricultural employment and GVA supports increases in non-agricultural activity, mainly in eastern Europe, it does not follow that rural development should be focused on the agricultural sector. These are also the regions where there is the most urgent need for structural change in agriculture, and this is best supported by the creation of outside employment options for agricultural labour.
Broad-based rural policies which reduce access costs, improve human capital and support innovation are more likely to be relevant in all regions rather than sector policies which are likely to be neither efficient nor effective.
Photo credit David Hallam-Jones and used under a Creative Commons licence