Exploring the labour productivity of agricultural systems across European regions: A multilevel approach

Land Use Policy 77 (2018) 94–106

Contents lists available at ScienceDirect

Land Use Policy journal homepage: www.elsevier.com/locate/landusepol

Exploring the labour productivity of agricultural systems across European regions: A multilevel approach

T

⁎

Elias Giannakis , Adriana Bruggeman Energy, Environment and Water Research Center, The Cyprus Institute, 20 Konstantinou Kavafi Street, 2121, Nicosia, Cyprus

A R T I C LE I N FO

A B S T R A C T

Keywords: Agricultural labour productivity Multilevel logistic regression model Technical efficiency Soil erosion Farm education Water limited yield potential

Agricultural productivity varies greatly among agricultural systems and among regions in Europe. A multilevel logistic regression model was applied to investigate the labour productivity of the six main agricultural systems across European NUTS2 regions. K-means and two-step clustering methods were used to classify European regions based on the agricultural systems’ standard output per annual work unit. We analysed the effect of environmental (soil erosion, rainfed yield potential), structural (farm education, age, pluriactivity, diversification, rented agricultural land), technical (yield ratio) and contextual (gross domestic product per capita, population density) factors on labor productivity. Significant differences were revealed between northern-central regions and the continental peripheries (Mediterranean, Eastern). Soil erosion negatively affects agricultural labour productivity; for each one ton/ha increase in the modelled annual soil erosion rate the odds of regions to attain high labour productivity decreased by 28%. The importance of technical efficiency in crop production was also identified. Observed low wheat yields, relative to modelled potential yields, in the southern, eastern and northern European regions indicate a large unexploited gap. The positive effect of the regional gross domestic product per capita and the low population density confirmed the importance of contextual factors on labour productivity. A second analysis of a composite indicator of the labour productivity of the European agricultural systems, which accounted for the productivity of each system without considering its size, revealed the positive effect of farm education and the negative effect of pluriactivity on agricultural labour productivity. The analyses indicate the importance of compensatory allowances for areas facing natural constraints, while investing in farm training schemes and advisory services could increase the adoption of new technologies and improve the performance of farmers in both economic and environmental terms. Finally, the significance of contextual factors indicates the importance of a better harmonisation of rural development policy with regional policy.

1. Introduction Agriculture is the main land user in the European Union (EU), accounting for 40% of its total land area, that is, 174.6 million hectare (Eurostat, 2017a). For several decades now, the number of farm holdings is continuously decreasing, while the farm size indicates a tendency towards larger holdings. The average farm size in the EU-27 increased from 11.5 ha in 2003 to 16.2 ha in 2013 (Eurostat, 2017a). Southern European countries (Spain, Italy, Greece, Portugal and Cyprus), i.e., those countries hit hardest by the recent economic crisis, showed relatively low rates of decline in the number of holdings between 2007 and 2013, e.g., Portugal -3.9% (Eurostat, 2017a), reflecting the ability of agriculture to form safety nets during periods of economic downturn (European Commission, 2013a ; Giannakis and Bruggeman, 2017a). On the contrary, eastern European countries exhibit the highest rates of decline in the number of farm holdings (e.g., Slovakia:

⁎

−65.8%) (Eurostat, 2017a) as a result of the process of privatization and redistribution of agricultural land. The total agricultural labour force in the EU-27, expressed in annual work units (AWU), which includes and accounts for part-time and seasonal work, shrank by 34% between 2003 and 2013 (Eurostat 2017a; 2018); the average AWU per farm holding decreased from 0.91 in 2003 to 0.88 in 2013 (Eurostat 2017a, 2018). The agricultural jobs that remained in the sector have, however, become more productive: the average EU standard output (SO) generated per annual work unit, which can serve as a proxy for agricultural labour productivity, increased from 24,101 Euro in 2007 to 34,830 Euro in 2013 (at current prices) (Eurostat, 2017a). The most impressive increases of agricultural labour productivity between 2007 and 2013 took place in Slovakia (158%), Bulgaria (123%) and Latvia (112%) (Eurostat, 2017a). The agricultural labour productivity varies greatly across Europe. Significant differences are revealed between the continental northern-

Corresponding author. E-mail addresses: [email protected] (E. Giannakis), [email protected] (A. Bruggeman).

https://doi.org/10.1016/j.landusepol.2018.05.037 Received 11 December 2017; Received in revised form 27 March 2018; Accepted 18 May 2018 0264-8377/ © 2018 Elsevier Ltd. All rights reserved.

Land Use Policy 77 (2018) 94–106

E. Giannakis, A. Bruggeman

advantages. However, there are no system-specific data at regional level that allow the quantification of such effects. Within this context, the objectives of this paper are: (a) to identify the differences in the labour productivity of agricultural systems across European countries and regions; (b) to investigate the factors that influence agricultural labour productivity; (c) to discuss the policy implications related with the significance and magnitude of those factors.

central countries and the continental peripheries, i.e., the Mediterranean, Eastern Europe and Scandinavia. For example, Denmark has the highest standard output per annual work unit, that is, 155,717 Euro/AWU (2007-2013) followed by Belgium (120,674 Euro/ AWU), while Romania (5958 Euro/AWU) and Bulgaria (6704 Euro/ AWU) have the lowest (Eurostat, 2017a). The magnitude of the territorial asymmetries in labour productivity across the EU is significantly greater in agriculture than in the secondary and tertiary sectors (Ezcurra et al., 2008), constituting an impediment in achieving the territorial cohesion objectives, as designated in the Lisbon Treaty (European Commission, 2008a). The enhancement of agricultural productivity has been an overarching objective of the Common Agricultural Policy (CAP) over time, through financial support of farm restructuring and modernization (Hennis, 2005), while the large differences in agricultural productivity levels across European countries and regions is a challenge that CAP has faced from its establishment. Agricultural productivity has been extensively studied at country level (Martín-Retortillo and Pinilla, 2015; Ball et al., 2001) and farm level (Helfand and Levine, 2004; Mugera et al., 2012), while significantly fewer studies focus on regional (Ezcurra et al., 2011; Smit et al., 2015) and local, i.e., municipal (Ženka et al., 2016) differences. Most studies at regional level explore how contrasting forces generate the observed “convergence” or “divergence” of agricultural productivity across regions (Esposti, 2011; Alexiadis et al., 2013). Relatively few studies have examined the labour productivity of individual farming systems at regional level and even fewer studies have employed a cross-regional analysis of the labour productivity of the different agricultural systems. For example, Veysset et al. (2015) studied the trend of labour productivity of suckler beef production systems in France and Manrique et al. (1999) explored the labour productivity of livestock farming systems in Pyrenees, Spain. Błażejczyk-Majka et al. (2011) analysed the labour productivity of field crop farms and mixed farms across 80 regions belonging to eleven old EU member states (EU15) and four new EU member states (Czech Republic, Hungary, Poland, Slovakia). Limited is, however, the knowledge about the European inter-regional distribution of the labour productivity across agricultural systems. A systems specific approach can allow a better understanding of the variability of productivity since it takes into account the well-known apples and oranges problem (Bernard and Jones, 1996). It is therefore of interest to investigate the labour productivity at both sectoral, that is, agricultural systems, and spatial, that is, regional level. In other words, to explore why labour or capital intensive agricultural systems yield high value added per labour unit in some regions, while in other regions they fail to do so. The Commission Regulation (EC) No 1242/2008 describes in detail the typology of European agricultural holdings (European Commission, 2008b). The Regulation identifies eight specialist types of farming, namely, field crops (i.e., cereals, rice, dried pulses and protein crops, potatoes, sugar beet, tobacco and cotton), horticulture (i.e., vegetables and flowers), permanent crops (i.e., fruit plantations, olive plantations and vineyards), grazing livestock (i.e., cattle, sheep and goats), granivores (i.e., pigs, poultry and rabbits), mixed cropping (i.e., field crops and permanent crops, field crops and horticulture, horticulture and permanent crops), mixed livestock holdings (i.e., grazing livestock and granivores) and mixed crops-livestock (i.e., permanent crops and grazing livestock, field crops and grazing livestock, field crops and granivores). A holding’s farming type is determined by the production system that makes up more than two-third of the total standard output of the holding. Although there is no unique set of physical, technical and human capital factors and conditions that enhance agricultural labour productivity, it is, however, possible to identify common patterns for attaining high economic output per labour unit in agriculture. A separate analysis of the determinants of the labour productivity of the European agricultural systems would be ideal to reveal sectors’ comparative

2. Methodology 2.1. Agricultural productivity data Within this study, we analyse the labour productivity of six agricultural systems: (a) field crops, (b) horticulture, (c) permanent crops, (d) grazing livestock, (e) granivores, and (f) mixed crop-livestock. The two subgroups of the mixed livestock holdings system, namely, the mixed livestock – mainly grazing livestock, and the mixed livestock – mainly granivores, were combined with the grazing livestock and the granivores farming systems, respectively. Additionally, the mixed cropping agricultural system was omitted from our analysis due to its small share in the EU-28 (3% in terms of SO and 5% in terms of AWU). The data used in the analysis of the labour productivity of agricultural systems of the NUTS2 regions are the official EU Farm Structural Surveys (FSS) data (Eurostat, 2017b). These surveys are carried out every three or four years as a sample survey (e.g., 2007 and 2013), and once in ten years as a census (e.g., 2010). The FSSs provide comparable and representative statistics across countries and regions and time and this information is used as a basis for decision making in the CAP. The representativeness and reliability of FSSs at the level of NUTS2 regions and by farm type is described in the Regulation (EC) No 1166/2008 (European Commission, 2008c). A summary of the farm structure statistics (standard output, annual work units, utilized agricultural area and livestock units) of the six farming systems and derived productivity indicators, at the EU and country level, is presented in Section 3.1. Correlations between the labour productivity and the agricultural area or livestock units of the agricultural systems at the country level are computed. 2.2. Explanatory variables The explanatory variables of the model can be classified into four broad categories, including farm structural factors (farmer and farm holding characteristics), environmental factors, technical factors and contextual factors. The variables were averaged across a six-year period (2007-2013) to mitigate potential year-specific effects in agricultural labour productivity caused by production and price fluctuations, except for the environmental and technical factors, as explained below. The reference spatial unit for the cross-regional empirical analysis is NUTS2 regions. In the case of Germany, we used the data of the NUTS1 regions due to constraints in data availability at NUTS2 level. Table 1 presents the definitions and the descriptive statistics of all variables used in the analysis. The first two factors, namely, age and farm education, represent two human capital aspects of the farm population. The age of the farm population is expressed as the share of farmers older than 55 years. The age (AGE) is commonly related to farmers’ managerial skills and aptitude to innovate, adopt modern farm practices and technologies and gain funding opportunities (Ezcurra et al., 2011). The farm education (FEDUC) is expressed as the share of farm managers with basic or full agricultural training. A better trained farm population facilitates the introduction of technical innovation, the absorption of externally generated knowledge and plays an important role in the adaptation of the sector to climate change (Vecchione, 2010; Giannakis et al., 2016). Farmers’ engagement in other gainful activity other than farm work (PLUR) is an important adaptation strategy to cope with market pressures (Weltin et al., 2017; Giannakis et al., 2018) and can have either 95

Land Use Policy 77 (2018) 94–106

E. Giannakis, A. Bruggeman

Table 1 Description of the 10 explanatory variables used in the multilevel logistic regression models and key summary statistics for the 238 European NUTS2 regions, 20072013. Variables

Definition

Source

Min

Max

EU average

AGE FEDUC PLUR HINDEX TENURE SOIL YIELDPOT YIELDRATIO

Share of farmers older than 55 years (%) Share of farm managers with basic or full agricultural training (%) Share of farm holders with other gainful activity other than farm work (%) Agricultural systems diversity in terms of annual work units (index) Share of rented agricultural land (%) Potential soil erosion by water (ton/ha/year)a Rainfed wheat yield potential (ton/ha)b Yield ratio: actual wheat yield (ton/ha)c / rainfed wheat yield potential (ton/ha)

1 1 5 0.18 7 0.05 1.6 0.1

84 99 86 0.85 90 18.07 9.8 2.0

51 41 38 0.34 42 2.70 7.3 0.8

POPDENS RGDP

Population density (100 persons per km2) Regional gross domestic product at current market prices per capita (1000 Euro/inhabitant)

European Commission (2017a) European Commission (2017b) European Commission (2017c) Eurostat (2017c) Eurostat (2017d) Eurostat (2017e) MARS (2017) Eurostat (2017f); DEFRA (2017); MARS (2017) Eurostat (2017g) Eurostat (2017h)

0.03 3.4

69.8 79.8

3.3 24.5

a b c

Soil loss by water erosion in European regions was estimated for the reference year 2010 (see Panagos et al., 2015). Rainfed wheat yield potential is an average of 2000, 2010 and 2015 (MARS, 2017). Actual wheat yield is the average of the regional yield statistics for the years 2000–2015.

yield of these three years deviated less than 10% from the 2000–2015 average (Eurostat, 2017f, DEFRA, 2017). Thus, we assumed that the weather conditions of these three years, and hence the simulated rainfed yield potential, could be considered representative for the wheat-growing conditions of the 2000–2015 period. The ratio between the observed, actual wheat yield and the simulated potential rainfed wheat yield is used as a proxy for the technical efficiency in crop production (YIELDRATIO). We used the average of the wheat yields reported in the statistics for 2000–2015 (Eurostat, 2017f, DEFRA, 2017) as the numerator and YIELDPOT as the denominator for the ratio. The population density (POPDENS) and the regional per capita output (RGDP) represent two contextual factors that are expected to affect labour productivity in agriculture. The population density is commonly used as a proxy for urbanization economies and access to consumers (Ženka et al., 2016). Agricultural areas located near urban centers, which are the main recipients of agricultural products, are more intensively cultivated since the proximity to the marketplace ensures a demand for the agricultural production (Polyzos and Arambatzis, 2006). However, other studies found that population density has a negative effect on agricultural labour productivity (Smit et al., 2015). The regional gross domestic product per capita, a proxy of the level of regional development, has been found to affect the agricultural labour productivity positively (Ezcurra et al., 2011). The institutional, legal and governance environment in which farms are embedded also largely influence the labour productivity in agriculture (Barro and Sala-I-Martin, 2004). However, there are no data at regional level that allow the quantification of such effects.

positive or negative effect on agricultural labour productivity since it may foster innovation through the interaction with other sectors (Schmitt, 1988; OECD, 2013) or on the contrary, promote structural change in the farm sector, which consequently leads to future exit from farming (Weiss, 1999; Kimhi, 2000). A Herfindahl index (HINDEX) is used to measure agricultural systems employment diversity in terms of annual work units based on the following formula: n

Hi =

Eij ⎞2

∑ ⎛E ⎜

j=1

⎝

i

⎟

⎠

(1)

where Hi is the value of the Herfindahl index in region i ; Eij is the employment in region i in agricultural system j ; Ei is the total agricultural employment in region i ; and n the number of agricultural systems (n = 6) . High values of the Hi index (close to 1) signify concentration of employment in fewer agricultural systems. The share of rented agricultural land (TENURE) is used to assess the effect of ‘non-owned’ factors of production on labour productivity. Llewelyn and Williams (1996) reported that farmers who own their land exhibit higher efficiency due to better resource management practices. On the contrary, Karagiannis and Sarris (2005) found that for the case of Greek tobacco growers the increase of the portion of rented land enhanced technical efficiency and productivity. Soil erosion by water (SOIL), one of the major soil threats in Europe (Panagos et al., 2015), is one of the indicators in this study that captures the effect of unfavourable environmental conditions on agricultural labour productivity. Potential soil erosion, henceforth referred to as soil erosion, was estimated by the Revised Universal Soil Loss Equation model (RUSLE2015), computed by Panagos et al. (2015). The highest soil loss rates are found in Mediterranean regions and in the Alpine regions of Slovenia and western Austria mainly due to high rainfall erosivity and steep topography (Panagos et al., 2015). The other environmental indicator is the simulated, rainfed wheat yield potential (YIELDPOT). Wheat, a predominantly rainfed crop, is the most widely grown cereal in Europe. The rainfed wheat yield potential, also referred to as water-limited yield potential, is defined as the maximum yield that can be obtained when wheat is grown in rainfed conditions, that is, limited by water supply and hence influenced by soil type and field topography (Boogaard et al., 2013; van Ittersum et al., 2013). Simulated rainfed wheat yield potential data for the years 2000, 2010 and 2015 were available from the Monitoring Agricultural Resources (MARS) Unit of the Joint Research Center (JRC) of the European Commission (MARS, 2017). To ensure that these three years were not marked by exceptional dry or wet conditions that affect wheat production, we analysed the reported observed yield series for the full 16-year period. For the majority of the regions, the average observed

2.3. Cluster analysis European regions are classified in clusters according to the labour productivity of the total agricultural sector and, secondly, according to the labour productivity of the six different agricultural systems. Cluster analysis entails various approaches and algorithms for grouping the observations, namely, hierarchical methods, partitioning methods, and two-step clustering (Hair et al., 2006). The k-means method and the two-step clustering procedure were used in this study to cluster European regions in homogeneous groups. The k-means clustering method splits the observations into a prearranged number of k homogeneous clusters by minimizing the within-cluster variation (Lattin et al., 2003). The two-step clustering, which is a combination of hierarchical and partitioning methods, assigns the observations into pre-clusters in the first stage, while in the second stage a modified hierarchical algorithm is used to group the pre-clusters (Chiu et al., 2001). The high and low productivity NUTS2 regions, for the whole sector and per system, are mapped and described in Section 3.2. 96

Land Use Policy 77 (2018) 94–106

E. Giannakis, A. Bruggeman

Table 2 Key indicators (annual average 2007–2013) for the European agricultural sector as a whole and for the six European farming systems, where SO is Standard Output (Euro), AWU is Annual Work Units, UAA is Utilized Agricultural Area (ha), LSU is Livestock Units (Source: Eurostat, 2017b). Indicator

Total agricultural sector

Field crops

Horticulture

Permanent crops

Grazing livestock

Granivores

Mixed crop-livestock

SO AWU Holdings UAA LSU SO/AWU SO/holding SO/UAA AWU/holding AWU/UAA UAA/holding UAA/total area LSU/holding LSU/total area

308.25 × 109 10.43 × 106 12.30 × 106 174.60 × 106 134.06 × 106 29,545 25,066 1,765 0.85 0.06 14.2 0.40 10.9 0.31

% 18 20 26 39 2 27,139 17,391 813 0.64 0.03 21.4 0.16 0.7 0.01

% 10 6 2 1 0 51,334 124,457 22,986 2.42 0.45 5.4 0.00 0.6 0.00

% 13 15 18 6 0 25,005 17,273 3,503 0.69 0.14 4.9 0.03 0.1 0.00

% 29 30 21 34 51 28,399 34,608 1,503 1.22 0.05 23.0 0.14 26.3 0.16

% 19 9 12 4 37 64,259 40,875 8,448 0.64 0.13 4.8 0.02 34.5 0.11

% 9 16 14 12 10 16,530 15,356 1,291 0.93 0.08 11.9 0.05 7.6 0.03

The odds ratios of the random intercept logistic regression model are used in the analysis of the results, as the coefficients of the regression model indicate the direction of the effect but not the size (Gelman and Hill, 2006). The focus here is to quantify the effect of the identified determinants (Section 2.2) on the probability of a region to attain high agricultural labour productivity (dependent variable) and not to provide a causal interpretation of those factors on the formulation of the agricultural labour productivity. Multicollinearity was tested and quantified using the variance inflation factor (VIF). A VIF higher than 5 implies that the regressions coefficients are poorly estimated because of multicollinearity (Montgomery et al., 2012). Statistical significance is assumed at the 10% level. All calculations were done in STATA 12 econometric software package.

2.4. Multilevel logistic regression analysis Regional employment and economic output dynamics are significantly affected by country-level effects (Ezcurra and Rapún, 2006; Giannakis and Bruggeman, 2017b). The empirical framework of the study is based on a multilevel model that distinguishes two levels of analysis, i.e., the regional level and the country level. A two-level random intercept model that relates a dichotomous variable y ∈ {0,1} , that is high/low agricultural labour productivity, to a set of explanatory variables (x1, …, x n ) , such as structural factors, environmental factors and contextual factors, can be formulated as follows:

yij = β0 + β1 x ij + uj + eij

(2)

where eij is the regional (level 1) residual and uj is the country (level 2) residual. The proportion of the residual variance in regional agricultural labour productivity that is attributed to between-countries variation is estimated by the variance partition coefficient (VPC) as follows (Goldstein et al., 2002):

VPC =

σu2

σu2

σu2 + σe2

3. Results 3.1. Labour productivity of agricultural systems across European countries A summary of key indicators for the six European agricultural systems is presented in Table 2. The grazing livestock system has the greatest contribution in the formulation of the total agricultural standard output (29%) and annual work units (30%), while the field crops system has the highest contribution in terms of the number of farm holdings (26%) and utilized agricultural area (39%) (Eurostat, 2017b). Between 2007 and 2013, the utilized agricultural area of field crops increased by 20%, whereas the area of all other systems decreased and the total agricultural area remained stable. During this period, the highest proportion of utilized agricultural area to total land cover, including grassland, was reported in UK (68%) and Ireland (67%), while the lowest proportion was reported in Finland (7%) and Sweden (7%) (Eurostat, 2017b). The granivores system attains the highest labour productivity, that is, standard output per annual work unit (64,259 Euro), followed by the horticulture (51,334 Euro) and the grazing livestock (28,399 Euro) systems, while the lowest levels of labour productivity are presented in the field crops (27,139 Euro) and the mixed crop-livestock system (16,530 Euro) (Eurostat, 2017b). The horticulture system attains the highest annual standard output per farm (124,457 Euro/holding) and standard output per hectare, that is, land productivity (22,986 Euro), while the mixed crop-livestock system exhibits the lowest standard output per farm (15,356 Euro/holding) and the arable crops system the lowest land productivity (813 Euro/ha). The horticulture system has the highest labour intensity per holding and per hectare (2.42 AWU/holding and 0.45 AWU/ha), while the arable crops system exhibits the lowest labour intensities (0.64 AWU/holding and 0.03 AWU/ha) (Eurostat, 2017b). The grazing livestock system, which exploits permanent grasslands, has the largest size in terms of utilized

(3)

σe2

where is the country-level residual variance and is the regionallevel residual variance. The logistic distribution of the regional residual (eij ) implies a variance equal to π 2/3, that is, 3.29 (Goldstein et al., 2002). Considering the different scales of operations, investments and production costs of the six agricultural systems, two analyses of the determinants of agricultural labour productivity are conducted within this study. The dependent variable of the logistic regression model in our first analysis is the labour productivity of the whole agricultural sector. The dependent variable of the model in our second analysis is a composite indicator of the labour productivity of the six agricultural systems that accounts for the productivity of each system without considering its size, thus representing a more diverse agricultural economy. The differences in the findings of these two analyses facilitate the identification of poor performing regions with productive smaller systems, and vice versa, and the factors that affect the inconsistent performance of regions. The composite indicator is a binary variable that takes the value 1, that is, high agricultural labour productivity, if the average of the dichotomous response variables of the six agricultural labour productivities is higher than 0.5, and 0 otherwise. Six cluster analyses were performed to group the regions according to the labour productivity of each agricultural system. Thus, a region with three high performing agricultural systems is classified as low performing and a region with four high performing systems is classified as high performing. 97

Land Use Policy 77 (2018) 94–106

E. Giannakis, A. Bruggeman

Fig. 1. Annual average labour productivity of agricultural systems across European countries, 2007-2013.

technique were used for the abovementioned classification of agricultural systems. Both clustering techniques gave a two-cluster solution for the agricultural systems of field crops, horticulture and mixed croplivestock and a three-cluster solution for the systems of permanent crops, grazing livestock and granivores. Considering, however, the very low number of regions in the third cluster, ranging between 1 and 5 regions (e.g., BE10 - Région de Bruxelles), this was merged with the second cluster. The labour productivity of the field crop system ranges from an average of 164,877 Euro/AWU for the top 5% regions, which are mainly located in Belgium, The Netherlands and France, to an average of 4376 Euro/AWU for the bottom 5% regions, mainly located in Poland. For the horticulture system labour productivity ranges from an average of 217,282 Euro/AWU for the top 5% regions, which are mainly located in Ireland, Slovakia and Sweden, to 7828 Euro/AWU for the bottom 5%, mainly located in Bulgaria, Lithuania and Romania. The top 5% regions for the permanent crop system were mainly located in Denmark and the UK and had an average labour productivity of 231,968 Euro/AWU, while the bottom 5% were mainly located in Bulgaria and Hungary with an average labour productivity of 3134 Euro/AWU. For the grazing livestock system labour productivity ranges from an average of 130,040 Euro/AWU for the top 5% regions, which are mainly located in Denmark, The Netherlands and Belgium, to 4091 Euro/AWU for the bottom 5%, mainly located in Poland, Bulgaria and Romania. The labour productivity of the granivores system ranges from an average of 398,599 Euro/AWU for the top 5% regions, which are mainly located in Italy, to an average of 5266 Euro/AWU for the bottom 5% regions, mainly located in Romania and Bulgaria. The top 5% regions for the mixed-livestock system were mainly located in The Netherlands and Belgium and had an average labour productivity of 163,476 Euro/AWU, while the bottom 5% were mainly located in Bulgaria, Romania and Portugal with an average labour productivity of 3201 Euro/AWU. The northern-central part of Europe, consisting of Danish, Belgian (with the exemption of arable crops system in BE34 - Prov. Luxembourg and horticulture and mixed crop-livestock systems in BE10 - Région de Bruxelles), Dutch (with the exemption of arable crops system in NL22 Gelderland and NL31 – Utrecht and permanent crops system in NL12 Friesland) and the majority of German regions, is characterised by high labour productivity of all agricultural systems. On the contrary, all eastern European regions of Estonia, Latvia, Lithuania, Poland, Hungary, Romania and Bulgaria exhibit low labour productivity rates.

agricultural area (23 ha/holding), while the granivores system has the highest number of livestock units (34.5 LSU/holding). Country-level labour productivities of the grazing livestock, granivores and mixed crop-livestock systems are positively associated with the number of livestock units per holding (Pearson correlation coefficient R: 0.79, 0.71 and 0.70, respectively). For the crop-based systems, a positive correlation is found between labour productivity and utilized agricultural area per holding for the horticulture system (R: 0.58), while weak positive associations are found for the permanent crops system (R: 0.42) and for the field crops system (R: 0.35). Fig. 1 presents the uneven distribution of the labour productivity of the six agricultural systems across European countries. Significant disparities in the value added per labour unit are presented both between agricultural systems and across countries, i.e., at both sectoral and spatial level. For example, labour productivity ranges from 391,413 Euro/AWU (Italy) to 5333 Euro/AWU (Romania) for the granivores system, from 237,494 Euro/AWU (Denmark) to 2923 Euro/AWU (Bulgaria) for the permanent crops system, from 163,768 Euro/AWU (Ireland) to 6023 Euro/AWU (Lithuania) for the horticulture system, from 144,957 Euro/AWU (Denmark) to 3738 Euro/AWU (Bulgaria) for the grazing livestock system, and from 103,596 Euro/AWU (Belgium) to 6909 Euro/AWU (Poland) for the field crops system. As we previously mentioned, the granivores system and the horticulture system attain the highest labour productivity at European level (Table 2). The share of the granivores system in the formulation of the total standard output and total annual work units is 19% and 9%, respectively, while the share of the horticulture system in the formulation of the total standard output and total annual work units is 10% and 6%, respectively (Table 2). In Italy, the share of the granivores system in the formulation of the total standard output and annual work units is 20% and 2.2%, respectively, which partially explains the high labour productivity of the sector in the country. Similarly, the highest labour productivity of the horticulture system is observed in Ireland and Slovakia, where the contribution of the sector in the formulation of the total standard output is 3.2% and 6.3%, respectively, and in the formulation of the total annual work units is 0.6% and 1%.

3.2. Labour productivity of agricultural systems across European regions The spatial distribution of the high and low labour productivity clusters of the six agricultural systems across the European regions is displayed in Fig. 2. The two-step clustering technique and the k-means 98

Land Use Policy 77 (2018) 94–106

E. Giannakis, A. Bruggeman

Fig. 2. High and low labour productivity of agricultural systems across European regions, 2007-2013, as identified by cluster analysis.

99

Land Use Policy 77 (2018) 94–106

E. Giannakis, A. Bruggeman

high labour productivity and one region had a dominant field crops system with high labour productivity. The factors that affect the underperformance of these regional agricultural economies are identified in the next section.

The horticulture is the only farming system that manages to attain high labour productivity in Slovak regions. Although the climate and the northern location of the Eastern-Scandinavian regions, i.e., Swedish and Finnish, is expected to negatively affect the competitiveness of agriculture, grazing livestock and horticulture systems attained high labour productivity. The most interesting differentiation among agricultural systems takes place in the southern part of Europe. The Spanish regions exhibit low performance in terms of standard output per labour unit in the permanent crops system but for the other five agricultural systems, especially the grazing livestock and the granivores, the majority of regions exhibit high labour productivity levels. This can be explained by the relatively large size of the livestock holdings in the Spanish regions, which as shown in Section 3.1 is positively associated with labour productivity. On the contrary, almost all Greek (with the exemption of horticulture system in EL13 - Dytiki Makedonia) and Portuguese (with the exemption of arable crops system in PT18 - Alentejo) regions are characterised by low labour productivity levels in all six agricultural systems. Alentejo is the largest agricultural region of Portugal, which produces 82% of the country’s wheat and is characterized by large farms with an average standard output per farm of 34,824 Euro (20072013) compared to a standard output per farm of 15,187 Euro at national level (Eurostat, 2017a; 2017f). A rapidly growing horticulture sector is observed in Dytiki Makedonia, which managed to (a) increase the utilized agricultural area of the sector by 238% in the period 2007–2013 (compared to an average reduction of 30% at national level), (b) increase the standard output of the sector by 175% (-32% at national level), (c) increase the annual work units of the sector by 7% (-32% at national level) (Eurostat, 2017b). In France, the field crops, permanent crops and mixed crop-livestock systems exhibit the highest labour productivity at regional level, while horticulture exhibit low labour productivity rates in the majority of regions. Similarly, in Italy, the mixed crop-livestock and permanent crops system exhibit very low performance, while on the contrary the granivores system attain high labour productivity rates in 15 out of 21 regions.

3.4. Factors behind the variability of agricultural labour productivity across European regions The results of the multilevel logistic regression models for the labour productivity of the agricultural sector and for the composite labour productivity indicator for the six agricultural systems across European regions are presented in Table 3. No multicollinearity was detected among the independent variables. The VPC of the multilevel logistic regression model for the agricultural sector equals 0.56, which indicates that 56% of the variance in the probability of a regional agricultural sector to attain high labour productivity is attributable to between-country effects. If we analyse the number of productive sectors (composite indicator) instead of the total size of the agricultural sector, the country effect becomes less prominent (VPC 0.43). Our empirical analysis for the whole agricultural sector confirms the negative effect of unfavourable environmental conditions on labour productivity in agriculture. An increase in the annual soil erosion by water (SOIL) by 1 ton/ha decreases regions’ probability of attaining high agricultural labour productivity by 28%. On the contrary, the positive effect of the rainfed wheat yield potential (YIELDPOT) on agricultural labour productivity is not statistically significant. Our model results confirm the positive effect of the technical efficiency of crop production (YIELDRATIO), represented here by the wheat yield ratio, on the agricultural labour productivity. The low technical efficiencies occur mainly in the southern, eastern and northern European regions (see Fig. 4), due to the low observed yields, thus indicating a large unexploited gap. Both contextual factors identified within this study, that is, population density and level of economic development, have statistically significant influence on the probability of attaining high labour productivity. More precisely, a hundred persons per km2 increase in population density decreases the probability of regions attaining high labour productivity by 14%. On the contrary, one thousand Euro increase in regional gross domestic product per capita increases the probability of attaining high labour productivity by 21%. The low percapita economic output regions are mainly located in the central and eastern European regions (e.g., BG31 – Severozapaden: 3383 Euro) (Fig. 4). Similar as the results of the multilevel logistic regression model for the regional labour productivity of the whole agricultural sector, soil erosion (SOIL), wheat yield ratio (YIELDRATIO), population density (POPDENS) and regional gross domestic product per capita (RGDP) are statistically significant determinants for the composite indicator of the six agricultural systems (Table 3). However, the analysis of the composite indicator indicated that the positive effect of farm education (FEDUC) in agricultural labour productivity is now statistically significant. A one percent increase in the share of farmers with basic or full agricultural training increases the odds of regions to attain high agricultural labour productivity by 5%. Similarly, the negative effect of pluriactivity (PLUR) in agricultural labour productivity becomes now statistically significant. The statistical significance of the FEDUC and PLUR variables in the composite indicator analysis can be attributed to the following changes: the regions that clustered as low labour performing regions in the agricultural sector analysis but as high labour performing regions in the composite indicator analysis (Fig. 3) were mainly characterized by high shares of farmers with agricultural training (e.g., Italian and French regions), while the regions that shifted from the high labour performing sector productivity cluster to the low labour performing composite indicator cluster, were mainly characterized by high shares of pluriactive farmers (e.g., Finnish regions).

3.3. Regional agricultural labour productivity clusters The two-step clustering technique gave a two-cluster solution for the productivity indicator of the whole agricultural sector. The k-means technique also gave the same classification of clusters. The high agricultural labour productivity cluster accounts for 118 regions with an average standard output of 83,022 Euro per annual work unit. The low agricultural productivity cluster accounts for 120 regions with an average standard output of 22,197 Euro per annual work unit. The DK05 - Nordjylland region has the highest standard output per annual work unit (173,790 Euro), while BG41 - Yugozapaden (3653 Euro/ AWU) is at the other extreme. The composite indicator, which was constructed from the classifications of the six systems (Section 3.2), was found to hold 120 high productive and 118 low productive regions. The distribution of agricultural labour productivity across European regions for the total agricultural sector (left) and the composite indicator (right) is displayed in Fig. 3. The results of the cluster analysis revealed that for 22 out of 238 regions the value of the composite indicator of agricultural labour productivity is different from that of the whole sector indicator. Eight out of the twelve regions that were identified as having low agricultural labour productivity in the whole sector analysis but high labour productivity according to the composite indicator analysis, had a dominant sector (i.e., more than 30% of the total standard output) with low productivity. Five regions had a dominant grazing livestock system and three regions had a dominant permanent crops system. Vice-versa, five of the ten regions that were identified as having high labour productivity according to the whole sector indicator and low labour productivity according to the composite indicator had a dominant grazing livestock system with high labour productivity, two regions had a dominant permanent crops system with 100

Land Use Policy 77 (2018) 94–106

E. Giannakis, A. Bruggeman

Fig. 3. Agricultural labour productivity across European regions for the total agricultural sector and for the composite indicator, 2007-2013.

Spanish diet that is nowadays higher than the average value for western continental countries and equivalent to the percentage found in the USA (Lassaletta et al., 2014). Moreover, Spain has become the second largest pork exporter in the EU-28 (19%), having overtaken Denmark (16%) and being behind Germany (25%) in 2016 (European Commission, 2017d), while the growing trade in suckler cows is further pushing up labour productivity (European Commission, 2013b). De Rancourt et al. (2006) analysed the performance of sheep and goat production systems in Spain, France, Italy and Greece and found that the Spanish farms attain the highest profitability for the sheep system (25,000 Euro/AWU) in 2001, the Italian farms attain the highest profitability for the goat system (20,000 Euro/AWU), while the Greek farms attain the lowest profitability for both sheep (10,000 Euro/AWU) and goat (11,000 Euro/AWU) production systems. The granivores system exhibits the highest productivity levels in terms of economic output generation per annual work unit, EU-wide, but on the other hand, livestock production is a major source of pollutants. According to Leip et al. (2015), the EU-27 livestock sectors for the year 2004 generated: 2.3 Tg N/yr NH3 emissions, that is, 82% of total agricultural NH3 emissions; 861 Tg CO2eq/yr, that is, 81% of greenhouse gases emissions from agriculture; 4.4 Tg N/yr diffuse N losses, that is, 73% of agriculture-related water pollution; and 76% of the agriculture-related terrestrial biodiversity loss. The highest livestock densities (LSU/total ha of region) are observed in the Belgian (e.g., BE25 - Prov. West-Vlaanderen: 3.9) and Dutch (e.g., NL41 Noord-Brabant: 3.6) regions, which are also characterized by the largest N inputs and surpluses across Europe (Bos et al., 2013; Willeghems et al., 2016). The horticulture system, which attains the second highest labour productivity, is generally intensively fertilized and irrigated, thus being highly vulnerable to N losses and causing serious environmental degradation (Di and Cameron, 2002). Recent EU legislative acts formulate a framework for better integration of environment, productivity and efficiency considerations in agriculture. These includes EU Directive 2016/2284, which aims to reduce emissions of agriculture from 2020 onwards through a set of measures such as low-emission manure spreading techniques, low-emission manure storage systems and low-emission animal housing systems; and EU Regulation 1305/ 2013, which fosters the competitiveness of all types of agriculture across all regions through farm restructuring, modernisation and increasing market participation. Giannadaki et al. (2018) indicated that the reduction of agriculture emissions generates a large net economic

Table 3 Odds ratios and significance of determinants of agricultural labour productivity using a multilevel logistic regression model for the whole agricultural sector and for the composite indicator of the six agricultural systems. Agricultural sector Odds ratio

Std. Error

Mixed-effects (regional level) AGE 1.04 0.04 FEDUC 1.04 0.03 PLUR 0.98 0.03 HINDEX 0.97 0.02 TENURE 1.03 0.02 SOIL 0.72 0.10 YIELDPOT 1.28 0.43 YIELDRATIO 1.03 0.02 POPDENS 0.86 0.03 RGDP 1.21 0.06 Constant 0.00 0.00 Random-effects (country level) Constant 4.15 3.79 Number of obs. 238 Log likelihood −63.65 p-value 0.0001 VPC 0.56

Composite Indicator p

Odds ratio

Std. Error

p

0.327 0.190 0.527 0.156 0.174 0.023 0.458 0.094 0.000 0.000 0.015

1.02 1.05 0.96 0.98 1.02 0.70 1.42 1.05 0.84 1.15 0.00

0.03 0.03 0.02 0.02 0.02 0.10 0.44 0.02 0.04 0.05 0.00

0.538 0.078 0.074 0.349 0.273 0.013 0.257 0.004 0.000 0.001 0.017

2.47

2.10 238 −63.05 0.0000 0.43

4. Discussion The investigation of the labour productivity among the six agricultural systems across European regions revealed, apart from the differences between the northern-central regions and the continental peripheries (Mediterranean, Eastern), an interesting differentiation within southern European regions. Despite the environmental constraints, as indicated by high soil erosion rates, and the low technical efficiency in crop production, some southern continental regions manage to attain a comparative advantage in the production of specific agricultural systems, while others fail. The most typical case is for the Spanish regions, which attain high labour productivity in the grazing livestock and granivores production systems. During the last 50 years, the Spanish livestock sector has been highly intensified with a vast increase in livestock numbers, particularly pigs and poultry, which can partially be attributed to the growing share of animal protein in the 101

Land Use Policy 77 (2018) 94–106

E. Giannakis, A. Bruggeman

Fig. 4. Regional values, classified into four quartiles (very low to very high), of the six statistically significant determinants of agricultural labour productivity: SOIL, YIELDRATIO, RGDP, POPDENS, FEDUC, PLUR. See Table 2 for descriptions of variables.

102

Land Use Policy 77 (2018) 94–106

E. Giannakis, A. Bruggeman

regions. The low technical efficiency of eastern European regions can also be attributed to structural factors such as the large land fragmentation as a result of the land privatization reforms, and to the overall transition from centrally planned to market oriented economics (Hartvigsen, 2014; Levers et al., 2016). On the contrary, the northerncentral part of Europe, which is characterised by low soil loss rates by erosion, experiences high yield ratios. The German, Dutch and Danish farms are the most capital intensive in the EU-15, invest more and are the largest in terms of economic output, while on the contrary the Greek and Portuguese farms are the least capital intensive, invest the least and are the smallest in terms of output (Rizov et al., 2013). Considering the global slowdown in the investment in agricultural research and development and the reallocation of funds towards other concerns, such as environmental effects of agriculture, food quality and food security, a reinvestment in agricultural research and development is critical to ensure high rates of farm productivity growth in the coming decades (Alston et al., 2009). Although CAP has multiple environmental and societal policy targets, its main objective is still the delivery of an efficient agricultural production system capable of competing on world markets (Quiroga et al., 2017). Dwyer et al. (2008), in their review of rural development instruments, highlight the positive effect of public support to farm investments on labour productivity and technical efficiency. Smit et al. (2015) found a statistically significant positive relationship between the rural development programmes spending on farm competitiveness measures and the agricultural labour productivity for the southern European regions, but not for the northern regions. They also found that spending is most effective in regions with intermediate population density. Our analysis indicates that population density is an important determinant for attaining high labour productivity in agriculture. More precisely, agricultural labour productivity is higher in less densely populated regions than in areas with high population densities. Smit et al. (2015) also found a negative and statistically significant effect of population density on agricultural labour productivity on European regions. On the contrary, Ženka et al. (2016) found only weak positive and insignificant effect of urbanization economies on agricultural labour productivity across Czech municipalities. Agricultural areas close to large cities are usually specialised in the intensive production of high-value products, including vegetables, fruits and milk, to benefit from urban demand (Bowler, 2014). For example, the SK01 - Bratislavský kraj region, i.e., the most urbanised region of Slovakia, where the capital of the country is located, attains the highest labour productivity in horticulture. The model results confirm the hypothesis that the regional gross domestic product per capita, has a positive effect on agricultural labour productivity. Ezcurra et al. (2011) reported similar results when examining the spatial distribution of agricultural productivity across the European regions for the period 1990-2000. These findings may indicate the effect of European regional and agricultural policies, through the distribution of the structural funds, i.e., European Regional Development Fund and European Agricultural Fund for Rural Development (European Commission, 2015), on regional growth and consequently on agricultural labour productivity. However, a better coordination and integration of rural development policies and regional policies within a common strategic framework is necessary, while there is a significant room for improvement in the allocation mechanisms of those structural funds to increase their spatial concentration and target financial resources towards disadvantaged regions (Crescenzi et al., 2015). The positive effects of rural development policies are mainly emerging in the most advanced European regions, with comparatively smaller gains for the disadvantaged regions (Crescenzi and Giua, 2016). The analysis of the composite indicator, which accounts for the productivity of the six agricultural systems without considering their size, revealed the positive influence of farm education on the formulation of labour productivity in agriculture. Giannakis and Bruggeman (2015) found that the odds for agricultural sectors with high shares of farmers with formal agricultural training to attain high

benefit for the EU under selected abatement options (i.e., low nitrogen feed strategies, low emission animal housing, manure storage capacity, urea application or substitution), the largest of which results from the introduction of highly efficient techniques for manure storage, that is, 163 billion US$. Model results indicate that the significance and the magnitude of the determinants of labour productivity in agriculture are strongly affected by country effects. These findings confirm the importance of applying multilevel modelling when data are hierarchically or nested structured, that is, units (e.g., regions) grouped at different levels of hierarchy (e.g., countries) (Gelman and Hill, 2006; Giannakis et al., 2018). Statistical methods that ignore the nesting structure of the data may result in underestimation of the variance of the estimated coefficients as well as in inconsistent parameter estimates (Rodriguez and Goldman, 1995). By applying multilevel analysis, we are able to account for and quantify the magnitude of country effects on regional agricultural systems productivity. Considering the large differences of agricultural labour productivity across European regions, integrated place-based rural development strategies, which take into account countries’ specificities, are required to reduce these disparities. Our findings confirm the negative influence of soil erosion on agricultural labour productivity. Bakker et al. (2007) found that wheat yields in northern Europe are not likely to be significantly reduced by soil erosion in the next 100 years, but for the southern countries the threat of erosion on crop productivity is clearly manifest. Soils play a key role in terrestrial ecosystems and the sustainability of agricultural production, thus minimizing soil degradation is essential for maintaining the services provided by soils (Schwilch et al., 2016). Panagos et al. (2015) suggest that soil protection measures should focus on the 24% of European land that experiences mean annual soil loss rates of over 2 ton/ha. Their modelling analysis demonstrated that a special focus should be given to arable land, where the implementation of the Good Agricultural and Environmental Conditions (GAEC) requirements introduced by the mid-term review of the CAP in 2003 has reduced the soil loss rate by 20%, resulting in an overall reduction of European soil loss by 9.5% during the period 2003–2010 (Panagos et al., 2015). The implemented measures include the promotion of reduced tillage and minimum soil cover, contour farming in sloping areas, the maintenance of terraces and stone walls, and the increased use of grass margins. These findings underline the importance of maintaining payments to farmers in mountain regions, the areas that suffer most from soil erosion, and other areas facing natural constraints, i.e., those areas in past referred as ‘less-favoured areas’, for encouraging continued use of agricultural land, preventing land abandonment and promoting sustainable farming systems (Regulation (EU) No 1305/2013). The findings of the analysis highlight the positive influence of the wheat yield ratio, a proxy for technical efficiency in crop production, on labour productivity in agriculture. Some regions have much greater potential than others to attain high yields due to their favourable climate and soil quality. However, not all of these regions managed to attain high yields indicating a large exploitable gap between current and potential yields, i.e., those theoretically achievable under best crop and land management practices (van Ittersum et al., 2013). Our yield ratios for the European regions (Fig. 4) showed a similar pattern as the yield gaps for the 1990–2006 period computed by Boogaard et al. (2013). These authors used actual yields from the Farm Accountancy Data Network and simulated potential yields with the MARS crop modelling system (the same system from which we retrieved our potential yield data). The most significant management gaps between actual and potential yields are observed in the southern continental regions, the Baltic regions and Finland. For example, the majority of the Portuguese regions, which are characterised by favourable conditions for rainfed crop production (high YIELDPOT) exhibit very low yield ratios. The eastern European regions also exhibit significant exploitable gaps between observed and potential wheat yields. Mueller et al. (2012) found insufficient nutrient inputs for wheat in Eastern European 103

Land Use Policy 77 (2018) 94–106

E. Giannakis, A. Bruggeman

improving the skills of farmers, fostering the uptake of innovations and enhancing farmers’ adaptation to a changing environment. Moreover, the results of the study highlight the significance of the regional economic development (measured by the gross domestic product per capita) and the low population density on shaping high labour productivity in agriculture, thus stressing the need for better harmonisation and integration of rural development policy with regional policy. Future research could analyse the impact of these different but strongly interrelated policies on the productivity of the European agricultural systems.

economic performance (expressed by the agricultural gross value added per farm holding) are 9 times greater than sectors with low levels of professional training across European countries. Farm training schemes and advisory services including issues such as climate change adaptation, can improve the knowledge and skills of farmers, increase the adoption of new technologies and innovations, and increase the climate resilience of farm holdings. Agricultural policies need to seek for a stronger weight to environmental protection and climate change considerations, compared to productivity and competitiveness goals (OECD, 2017). The analysis of the composite indicator revealed a negative association of pluriactivity with labour productivity in agriculture. Weiss (1999) investigated the structural change in the agricultural sector in Upper Austria for the period 1980–1990, and found that off-farm employment negatively affect farm growth and survival. Weltin et al. (2017) conducted a survey of 2154 farms in eleven case study regions in nine European countries to explore the differences among farm types regarding their willingness to diversify their income basis towards non-agricultural activities. They found that intensive livestock professionals were least likely to diversify their activities. Giannakis et al. (2018) investigated the determinants of off-farm employment in Cyprus and found that the operators of crop farming holdings have a 4.2-times higher probability to work off-farm than operators of livestock and mixed-farming holdings. A holistic analysis of the productivity of the European agricultural systems, which would facilitate an integrated policy impact assessment, requires system-specific data at regional level. However, many structural indicators (e.g., gross fixed capital formation per agricultural sector) are not available for all regions for all years, while for others there are time lags between data collection and reporting phases (for more details on data constraints for CAP evaluation, see European Commission (2016). The Regulation (EC) 1166/2008 requires the estimates to be representative at the level of NUTS2 regions and by farm type (European Commission, 2008c). However, the accuracy of the estimates is affected by sampling and non-sampling errors. Therefore, more efforts are needed to improve the reliability and quality of data and ensure uniform conditions across member states. For example, Hill (2012) mentions that in the EU Farm Structural Survey of 2000 less than one third of holdings kept accounts (e.g., in Greece 1% kept accounts and in Austria 5%). Environmental data such as water pollution are also difficult to analyse at sector and NUTS-region level. The scarcity of data for developing, evaluating and applying agricultural systems models is the main limitation of the current state of agricultural systems science (Jones et al., 2017). The establishment and maintenance of an integrated administration and control system (IACS) could optimise the integration of data and achieve transparency (Tóth and Kučas, 2016). The improved availability of data at sectoral and spatial level could allow the analysis of environmentally adjusted agricultural productivity, which measures the productivity growth of agricultural systems by taking into account the generation of undesirable environmental effects.

Acknowledgements The research has received financial support from the European Union Seventh Framework Programme under grant agreement No. 603498 (RECARE project). We would like to acknowledge the authors of the soil erosion and yield potential data for making their data available. References Alexiadis, S., Ladias, C., Hasanagas, N., 2013. A regional perspective of the common agricultural policy. Land Use Policy 30 (1), 665–669. Alston, J.M., Beddow, J.M., Pardey, P.G., 2009. Agricultural research, productivity, and food prices in the long run. Science 325, 1209–1210. Bakker, M.M., Govers, G., Jones, R.A., Rounsevell, M.D., 2007. The effect of soil erosion on Europe’s crop yields. Ecosystems 10 (7), 1209–1219. Ball, V.E., Bureau, J.C., Butault, J.P., Nehring, R., 2001. Levels of farm sector productivity: an international comparison. J. Productivity Analysis 15 (1), 5–29. Bernard, A.B., Jones, C.I., 1996. Comparing apples to oranges: productivity convergence and measurement across industries and countries. Am. Econ. Rev. 86 (5), 1216–1238. Barro, R.J., Sala-I-Martin, X., 2004. Economic Growth, 2nd ed. The MIT Press, Cambridge, England. Błażejczyk-Majka, L., Kala, R., Maciejewski, K., 2011. Productivity and efficiency of large and mixed farms of old and new EU regions. Agric. Econ.- Czech 58, 61–71. Boogaard, H., Wolf, J., Supit, I., Niemeyer, S., van Ittersum, M., 2013. A regional implementation of WOFOST for calculating yield gaps of autumn-sown wheat across the European Union. Field Crops Res. 143, 130–142. Bos, J.F., Smit, A.B.L., Schröder, J.J., 2013. Is agricultural intensification in the Netherlands running up to its limits? NJAS-Wageningen J. Life Sci. 66, 65–73. Bowler, I.R., 2014. The Geography of Agriculture in Developed Market Economies. Routledge, NY, USA, pp. 978–0582301610 ISBN-13. Chiu, T., Fang, D., Chen, J., Wang, Y., Jeris, C., 2001. A robust and scalable clustering algorithm for mixed type attributes in large database environment. Proceedings of the 7th ACM SIGKDD International Conference in Knowledge Discovery and Data Mining. Association for Computing Machinery, San Francisco, CA, pp. 263–268. Crescenzi, R., De Filippis, F., Pierangeli, F., 2015. In tandem for cohesion? Synergies and conflicts between regional and agricultural policies of the European Union. Regional Stud. 49 (4), 681–704. http://dx.doi.org/10.1080/00343404.2014.946401. Crescenzi, R., Giua, M., 2016. The EU cohesion policy in context: does a bottom-up approach work in all regions? Environ. Plann. A 48 (11), 2340–2357. http://dx.doi.org/ 10.1177/0308518X16658291. DEFRA, 2017. Regional Wheat Areas, Yields and Production 1999-2015. Available at:. Department for Environment Food & Rural Affairs, UK. https://data.gov.uk/dataset/ cereals_and_oilseeds_production_harvest. De Rancourt, M., Fois, N., Lavín, M.P., Tchakérian, E., Vallerand, F., 2006. Mediterranean sheep and goats production: an uncertain future. Small Ruminant Res. 62 (3), 167–179. Di, H.J., Cameron, K.C., 2002. Nitrate leaching in temperate agroecosystems: sources, factors and mitigating strategies. Nutrient Cycl. Agroecosyst. 64 (3), 237–256. Dwyer, J., Kirwan, J., Thomson, K., Clark, M., et al., 2008. DG Agri Project No. 2006-G410. Final Report. Review of Rural Development Instruments. University of Gloucestershire, UK. Esposti, R., 2011. Convergence and divergence in regional agricultural productivity growth: evidence from Italian regions, 1951-2002. Agric. Econ. 42 (2), 153–169. http://dx.doi.org/10.1111/j.1574-0862.2010.00508.x. European Commission, 2008a. Green Paper on Territorial Cohesion - Turning Territorial Diversity into Strength, Commission of the European Communities. Commission of the European Communities, SEC, pp. 2550. European Commission, 2008b. Commission regulation (EC) no 1242/2008 on establishing a Community typology for agricultural holdings. Off. J. Eur.Union L 335, 3–24. European Commission, 2008c. Regulation (EC) no 1166/2008 of the European parliament and of the council of 19 November 2008 on farm structure surveys and the survey on agricultural production methods and repealing council regulation (EEC) no 571/88. Off. J. Eur. Union L 321, 14–34. European Commission, 2013a. Structure and Dynamics of EU Farms: Changes, Trends and Policy Relevance. EU Agricultural Economics Briefs, No 9. European Commission, 2013b. EU beef farms report 2012 based on FADN data.

5. Conclusions This article has made an empirical exploration of the differences of labour productivity in agricultural systems across European regions. Model results indicate that technical efficiency, soil erosion, regional economic development and low population density are important determinants for agricultural labour productivity. The significance and the magnitude of the determinants of agricultural labour productivity are strongly affected by country effects as well as by the relative size of the agricultural systems. This reinforces the need to adopt an integrated approach for the analysis of this complex topic. Our findings highlight the importance of maintaining compensatory allowances for areas facing natural constraints to ensure sustainable agricultural production systems in these areas. Strengthening the training of European farmers and improving the efficiency of advisory services are pivotal for 104

Land Use Policy 77 (2018) 94–106

E. Giannakis, A. Bruggeman

Information Gap. Cabi. 4th ed. http://dx.doi.org/10.1079/9781845938475.0000. ISBN: 9781845938475. Jones, J.W., Antle, J.M., Basso, B., Boote, K.J., Conant, R.T., Foster, I., Godfray, H.C.J., Herrero, M., Howitt, R.E., Janssen, S., Keating, B.A., Munoz-Carpena, R., Porter, C.H., Rosenzweig, C., Wheeler, T.R., 2017. Toward a new generation of agricultural system data, models, and knowledge products: State of agricultural systems science. Agric. Syst. 155, 269–288. Karagiannis, G., Sarris, A., 2005. Measuring and explaining scale efficiency with the parametric approach: the case of Greek tobacco growers. Agric. Econ. 33 (s3), 441–451. Kimhi, A., 2000. Is part-time farming really a step in the way out of agriculture? Am. J. Agric. Econ. 82 (1), 38–48. Lassaletta, L., Billen, G., Romero, E., Garnier, J., Aguilera, E., 2014. How changes in diet and trade patterns have shaped the N cycle at the national scale: Spain (1961–2009). Region. Environ. Change 14 (2), 785–797. Analyzing multivariate data. In: Lattin, J.M., Carroll, J.D., Green, P.E. (Eds.), Duxbury Applied Series. Thomson Learning, Inc., Pacific Grove. Leip, A., Billen, G., Garnier, J., Grizzetti, B., Lassaletta, L., Reis, S., Simpson, D., Sutton, M.A., de Vries, W., Weiss, F., Westhoek, H., 2015. Impacts of European livestock production: nitrogen, sulphur, phosphorus and greenhouse gas emissions, land-use, water eutrophication and biodiversity. Environ. Res. Lett. 10 (11), 115004. Levers, C., Butsic, V., Verburg, P.H., Müller, D., Kuemmerle, T., 2016. Drivers of changes in agricultural intensity in Europe. Land Use Policy 58, 380–393. Llewelyn, R.V., Williams, J.R., 1996. Nonparametric analysis of technical, pure technical, and scale efficiencies for food crop production in East Java, Indonesia. Agricu. Econ. 15 (2), 113–126. Manrique, E., Olaizola, A., Bernués, A., Maza, M.T., Sáez, A., 1999. Economic diversity of farming systems and possibilities for structural adjustment in mountain livestock farms. Options Méditerranéennes 27, 81–94. MARS, 2017. Crop Yield Simulations on RCP. Agri4cast Resources Portal, Monitoring Agricultural ResourceS (MARS). Available at:. European Commission, Joint Research Centre. http://agri4cast.jrc.ec.europa.eu/DataPortal/ RequestDataResource.aspx?idResource=16&o=d. Martín-Retortillo, M., Pinilla, V., 2015. On the causes of economic growth in Europe: why did agricultural labour productivity not converge between 1950 and 2005? Cliometrica. 9 (3), 359–396. Introduction to linear regression analysis. New York, NY In: Montgomery, D.C., Peck, E.A., Vining, G.G. (Eds.), Wiley Series in Probability and Statistics, 5th ed. . Mueller, N.D., Gerber, J.S., Johnston, M., Ray, D.K., Ramankutty, N., Foley, J.A., 2012. Closing yield gaps through nutrient and water management. Nature 490, 254–257. http://dx.doi.org/10.1038/nature11420. Mugera, A.W., Langemeier, M.R., Featherstone, A.M., 2012. Labor productivity growth in the Kansas farm sector: a tripartite decomposition using a non-parametric approach. Agric. Resour. Econ. Rev. 41 (3), 298–312. OECD, 2013. Cross Country Analysis of Farm Performance. Organisation for Economic Cooperation and Development, Working Party on Agri-cultural Policies and Markets TAD/CA/APM/WP(2012)20/FINAL. OECD, 2017. Synthesis Report on Synergies and Trade-offs between Adaptation, Mitigation and Agricultural Productivity: Combining Qualitative and Quantitative Analysis. Organisation for Economic Cooperation and Development, Joint Working Party on Agriculture and the Environment COM/TAD/CA/ENV/EPOC(2017)6/ FINAL. Panagos, P., Borrelli, P., Poesen, J., Ballabio, C., Lugato, E., Meusburger, K., Montanarella, L., Alewell, C., 2015. The new assessment of soil loss by water erosion in Europe. Environ. Sci. Policy 54, 438–447. Polyzos, S., Arambatzis, G., 2006. Labour productivity of agricultural sector in Greece: determinant factors and interregional differences analysis. Mediterr. J. Econ. Agric Environ. 1, 58–64. Rodriguez, G., Goldman, N., 1995. An assessment of estimation procedures for multilevel models with binary responses. J. R. Stat. Soc. Ser. A 73–89. Quiroga, S., Suárez, C., Fernández-Haddad, Z., Philippidis, G., 2017. Levelling the playing field for European Union agriculture: does the common agricultural policy impact homogeneously on farm productivity and efficiency? Land Use Policy 68, 179–188. Rizov, M., Pokrivcak, J., Ciaian, P., 2013. CAP subsidies and productivity of the EU farms. J. Agric. Econ. 64 (3), 537–557. Schmitt, G.H., 1988. What do agricultural income and productivity measurements really mean? Agric. Econ. 2 (2), 139–157. Schwilch, G., Bernet, L., Fleskens, L., Giannakis, E., Leventon, J., Marañón, T., Mills, J., Short, C., Stolte, J., van Delden, H., Verzandvoort, S., 2016. Operationalizing ecosystem services for the mitigation of soil threats: a proposed framework. Ecol. Indic. 67, 586–597. Smit, M.J., van Leeuwen, E.S., Florax, R.J., de Groot, H.L., 2015. Rural development funding and agricultural labour productivity: a spatial analysis of the European Union at the NUTS2 level. Ecol. Indic. 59, 6–18. Tóth, K., Kučas, A., 2016. Spatial information in European agricultural data management. Requirements and interoperability supported by a domain model. Land Use Policy 57, 64–79. van Ittersum, M.K., Cassman, K.G., Grassini, P., Wolf, J., Tittonell, P., Hochman, Z., 2013. Yield gap analysis with local to global relevance - a review. Field Crops Res. 143, 4–17. Vecchione, G., 2010. EU Rural Policy: Proposal and Application of an Agricultural Sustainability Index. Munich Personal RePEc Archive Papers No. 27032. Available at:. http://mpra.ub.uni-muenchen.de/27032/. Veysset, P., Lherm, M., Roulenc, M., Troquier, C., Bébin, D., 2015. Productivity and technical efficiency of suckler beef production systems: trends for the period 1990 to 2012. Animal 9, 2050–2059.

Agriculture and Rural Development. Available at: http://ec.europa.eu/agriculture/ rica/pdf/beef_report_2012.pdf. European Commission, 2015. European structural and investment funds 2014-2020. Regional and Urban Policy. Publications Office of the European Union, Luxembourg. European Commission, 2016. CAP Context Indicators 2014-2020. Available at:. https:// ec.europa.eu/agriculture/sites/agriculture/files/cap-indicators/context/2016/fulltext_en.pdf. European Commission, 2017a. Age Structure of Farm Managers - Common Context Indicators. Available at:. https://ec.europa.eu/agriculture/sites/agriculture/files/ cap-indicators/context/2016/indicator-table_en.pdf. European Commission, 2017b. Agricultural Training of Farm Managers - Common Context Indicators. Available at:. https://ec.europa.eu/agriculture/sites/ agriculture/files/cap-indicators/context/2016/indicator-table_en.pdf. European Commission, 2017c. Farmers with Other Gainful Activity - Indicators in the Rural Development Report 2012 & 2013. Available at:. . https://ec.europa.eu/ agriculture/sites/agriculture/files/statistics/rural-development/2013/indicators_en. pdf. European Commission, 2017d. EU Meat Market Observatory – Pigmeat. Agriculture and Rural Development. Available at:. https://ec.europa.eu/agriculture/marketobservatory/meat/pigmeat/statistics_en. Eurostat, 2017a. Key Farm Variables: Area, Livestock (LSU), Labour Force and Standard Output (SO) by Agricultural Size of Farm (UAA), Legal Status of Holding and NUTS 2 Regions [ef_kvaareg]. Available at:. http://data.europa.eu/euodp/en/data/dataset/ CKFsL3lC9BK8zs6qtOMI1w. Eurostat, 2017b. Key Variables: Area, Livestock (LSU), Labour Force and Standard Output (SO) by Type of Farming (2-Digit) and NUTS 2 Regions [ef_kvftreg]. Available at:. http://appsso.eurostat.ec.europa.eu/nui/show.do?dataset=ef_kvftreg&lang=en. Eurostat, 2017c. Farm Indicators by Agricultural Area, Type of Farm, Standard Output, Legal Form and NUTS 2 Regions [ef_m_farmleg]. Available at:. http://appsso. eurostat.ec.europa.eu/nui/show.do?dataset=ef_m_farmleg&lang=en. Eurostat, 2017d. Type of Tenure: Number of Farms and Areas by Agricultural Size of Farm (UAA) and NUTS 2 Regions [ef_mptenure]. Available at:. http://appsso.eurostat.ec. europa.eu/nui/show.do?dataset=ef_mptenure&lang=en. Eurostat, 2017e. Soil Erosion by Water (tonnes per ha per year), EU-28. Available at:. http://ec.europa.eu/eurostat/statistics-explained/index.php/Agri-environmental_ indicator_-_soil_erosion. Eurostat, 2017f. Crop Statistics by NUTS 2 Regions (from 2000 Onwards) [agr_r_acs]. Available at:. http://ec.europa.eu/eurostat/web/products-datasets/product?code= agr_r_acs. Eurostat, 2017g. Population Density by NUTS 2 Region. Available at:. https://data. europa.eu/euodp/en/data/dataset/QEgn3fJF0SQo7qpAN8T9g. Eurostat, 2017h. Gross Domestic Product (GDP) at Current Market Prices by NUTS 2 Regions [nama_10r_2gdp]. Available at:. http://appsso.eurostat.ec.europa.eu/nui/ show.do?dataset=nama_10r_2gdp&lang=en. Eurostat, 2018. Key Variables by Legal Status of Holding, Size of Farm (UAA) and NUTS 2 Regions [ef_ov_kvaa]. Available at:. http://appsso.eurostat.ec.europa.eu/nui/show. do?dataset=ef_ov_kvaa&lang=en. Ezcurra, R., Rapún, M., 2006. Regional disparities and national development revisited: the case of Western Europe. Eur. Urban Region. Stud. 13 (4), 355–369. http://dx.doi. org/10.1177/0969776406068590. Ezcurra, R., Iraizoz, B., Pascual, P., Rapun, M., 2008. Spatial disparities in the European agriculture: a regional analysis. Appl. Econ. 40 (13), 1669–1684. Ezcurra, R., Iráizoz, B., Pascual, P., Rapún, M., 2011. Agricultural productivity in the European regions: trends and explanatory factors. Eur Urban Region. Stud. 18 (2), 113–135. Gelman, A., Hill, J. (Eds.), 2006. Data Analysis Using Regression and Multilevel/ Hierarchical Models, 1st ed. Cambridge University Press, New York, NY. Giannadaki, D., Giannakis, E., Pozzer, A., Lelieveld, J., 2018. Estimating health and economic benefits of reductions in air pollution from agriculture. Sci. Total Environ. 622, 1304–1316. Giannakis, E., Bruggeman, A., 2015. The highly variable economic performance of European agriculture. Land Use Policy 45, 26–35. http://dx.doi.org/10.1016/j. landusepol.2014.12.009. Giannakis, E., Bruggeman, A., Djuma, H., Kozyra, J., Hammer, J., 2016. Water pricing and irrigation across Europe: opportunities and constraints for adopting irrigation scheduling decision support systems. Water Sci. Technol. 16 (1), 245–252. Giannakis, E., Bruggeman, A., 2017a. Economic crisis and regional resilience: evidence from Greece. Papers Region. Sci. 96, 451–476. http://dx.doi.org/10.1111/pirs. 12206. Giannakis, E., Bruggeman, A., 2017b. Determinants of regional resilience to economic crisis: a European perspective. Eur. Plan. Stud. 25 (8), 1394–1415. http://dx.doi.org/ 10.1080/09654313.2017.1319464. Giannakis, E., Efstratoglou, S., Antoniades, A., 2018. Off-farm employment and economic crisis: evidence from Cyprus. Agriculture 8, 41. Goldstein, H., Browne, W.J., Rasbash, J., 2002. Partitioning variation in multilevel models. Underst. Stat. 1 (4), 223–231. http://dx.doi.org/10.1207/ S15328031US0104_02. Hair, J.F., Black, W.C., Babin, B.J., Anderson, R.E., Tatham, R.L. (Eds.), 2006. Multivariate Data Analysis, 6th ed. Pearson Prentice Hall, Upper Saddle River. Hartvigsen, M., 2014. Land reform and land fragmentation in Central and Eastern Europe. Land Use Policy 36, 330–341. Helfand, S.M., Levine, E.S., 2004. Farm size and the determinants of productive efficiency in the Brazilian Center-West. Agric. Econ. 31 (2–3), 241–249. Hennis, M., 2005. Globalization and European Integration: The Changing Role of Farmers in the Common Agricultural Policy. Rowman & Littlefield, Publishers. Hill, B., 2012. Farm Incomes, Wealth and Agricultural Policy Filling the CAP’s core

105

Land Use Policy 77 (2018) 94–106

E. Giannakis, A. Bruggeman

Willeghems, G., De Clercq, L., Michels, E., Meers, E., Buysse, J., 2016. Can spatial reallocation of livestock reduce the impact of GHG emissions? Agric. Syst. 149, 11–19. Ženka, J., Slach, O., Krtička, L., Žufan, P., 2016. Determinants of microregional agricultural labour productivity–Evidence from Czechia. Appl. Geogr. 71, 83–94.

Weltin, M., Zasada, I., Franke, C., Piorr, A., Raggi, M., Viaggi, D., 2017. Analysing behavioural differences of farm households: an example of income diversification strategies based on European farm survey data. Land Use Policy 62, 172–184. Weiss, C.R., 1999. Farm growth and survival: econometric evidence for individual farms in upper Austria. Am. J. Agric. Econ. 81 (1), 103–116.

106