SCALE EFFICIENCY IN EUROPEAN PORK PRODUCTION

The paper analyses scale efficiency in European pork production. The analysis shows significant differences in the exploitation of economies of scale among EU member states. In particular, old member states exhibit increasing returns to scale whereas most new member states show either constant or decreasing returns to scale. The differences among old and new member states are also pronounced from a dynamic perspective: whereas the old member states improved their productivity in pork production through scale efficiency, its impact in new member states was rather negative.


INTRODUCTION
The agricultural sector in new member states (NMS) of the European Union (EU) experienced significant changes in the period immediately preceding and after accession into the European Union.We can observe, among other things, changes in farm size.Since scale efficiency is an important factor in determining productivity change, the analysis of economies of scale has been a fundamental subject in agricultural economics in recent decades.
An optimal scale of production was defined by Frisch (1965) as an input bundle where scale elasticity equals one.That is, a farm operates under constant returns to scale.One can measure how close an observed farm is to the optimal scale using scale efficiency.According to Försund (1996) and Ray (1998), scale efficiency measures the average productivity at the observed input scale with respect to the optimal scale.Moreover, Coelli et al. (2005) consider scale efficiency to be a component of productivity change.
Studies on scale efficiency in European agriculture are numerous.2004 -2009) showed that livestock and mixed farms in NMS as well as old member states (OMS) enlarged their total agricultural area and animal stocks.However, scale efficiency increased only for mixed farms in the new regions.Our research complements current knowledge by comparing economies of scale and the contribution of scale efficiency to productivity changes in EU member states, for the time period 2004 -2011, by employing a translog stochastic meta-frontier model.In particular, the paper addresses questions related to the exploitation of economies of scale from a static as well as dynamic perspective.The aim is to evaluate differences, by country, in returns to scale and in the role played by scale efficiency in productivity changes in EU pork production.The paper is organized as follows: first, an estimation strategy and data set are described; next, the results of the stochastic metafrontier multiple output distance function estimate are presented and the estimated technology is discussed; and finally, the impact of scale efficiency on productivity change is analysed.

Theoretical framework
Productivity changes and the effects of economies of scale can be analysed by non-parametric or parametric methods.Kimura and Sauer (2015) represent the nonparametric approach based on the Fisher index.Bojnec and Latruffe (2013) used another non-parametric method: data envelopment analysis.An example of a parametric method is Piesse and Thirtle (2000), who applied stochastic frontier analysis.These methods have their pros and cons, e.g.Van Beren (2007) mentioned that the parametric method can lead to a bias in total factor productivity change, due to a selection bias connected with the construction of a balanced panel or with the endogeneity of inputs.This could be resolved by using a semi-parametric approach, which was used by Rizov et al. (2013), Mary (2013) and others.Our research is based on a parametric approach applied on unbalance panel data, and the endogeneity is resolved through random parameter model specification.
Specifically, the research questions will be addressed by the estimation of multiple output distance functions (ODF) for EU member states.To provide a coherent comparison of the contribution of scale efficiency to total factor productivity (TFP), we use the efficient outputs from the multiple output distance function country estimates in the estimation of a metafrontier multiple output distance function.Both the country and metafrontier models are formulated as translog multiple output distance functions with three outputs and five inputs: In all models we assume that agricultural production possibilities are significantly affected by firm heterogeneity, which affects the level as well as the shape of the production possibilities.The heterogeneity is captured by the fixed management model introduced by Alvarez et al. ( 2004).Moreover, it is assumed that the distribution is the same for all random parameters.That is, the model specification which is estimated in the empirical part (country-specific models as well as the metafrontier model) has the form: where m i * represents an observed heterogeneity.
( ) could possess any distribution with zero mean and unit variance and determines the level of the fitted parameters as well as the level of technical efficiency.
Furthermore, Alvarez et al. ( 2004) showed that u it can be estimated according to Jondrow et al. (1982).Total factor productivity is calculated in the form of the Törnqvist-Theil index (TTI) (Diwert, 1976 andCaves et al., 1982) and decomposed into its components: scale effect, technological change, technical efficiency and management effect (Cechura et al., 2014).All calculations are carried out using the econometric software NLOGIT 5.

Data
We use the Farm Accountancy Data Network (FADN) dataset provided by the European Commission.Every year, the FADN collect accountancy data from a sample of farms in the European Union.The survey covers farms which could be considered commercial, due to their size.The annual sample covers approximately 80,000 holdings, which represents approximately 90% of the total utilised agricultural area and accounts for about 90% of total agricultural production.
Our dataset consists of 24 EU member states (only Croatia, Cyprus, Luxemburg and Malta are missing) and covers the period from 2004 to 2011 (except for Austria (2005 -2011), Bulgaria andRomania (2008 -2011)).The variables we use in the analysis are as follows: y1 pork production, y2 other animal production, y3 plant production, x1 labour, x2 land, x3 capital, x4 specific material and x5 other material.
Labour is represented by the total labour, measured in average working unit (AWU).Land is the total utilised land.Capital is the sum of contract work and depreciation.Specific material in pork production represents feed costs for pigs and poultry.Outputs as well as inputs (except for labour and land) are deflated by country price indices on each individual output and input (2005 = 100).The country price indices are taken from the EUROSTAT database.
The multiple output distance functions are estimated only for specialized producers.Specialization is defined when pork production represents at least a 50 percent share of total animal production.Sample descriptive statistics, which are provided in Table 1, show that the major differences in the mean values of inputs (specific material is the exception) based on a country comparison are in the Czech Republic, Slovakia and Hungary.A higher mean value of plant production can also be observed in these countries compared to the rest.Farms in these countries have more diversified production.However, this also holds for the majority of new member states.

RESULTS AND DISCUSSION
First, we start with a general discussion of the fitted country-specific stochastic frontier models.Then, we will concentrate on a discussion of economies of scale.The complete results, together with a discussion, are provided in Cechura et al. ( 2014).
The majority of estimated parameters are significant at the 1% significance level.The parameter estimates of the multiple output distance function meet monotonicity requirements for all outputs and inputs.Land in Belgium, Great Britain and Slovakia, and capital in Bulgaria are exceptions.However, except for the case of Belgium, these parameters are not significant at the 5% significance level.In addition, the convexity of ODF in inputs holds for almost all countries evaluated on the sample mean.More about the properties of ODF can be found in Coelli et al. ( 2005).
The first-order parameters of outputs represent the share of outputs y2 (other animal production) and y3 (plant production) in total output.Pork producers with a higher share of other animal production can be found in the Czech Republic and Finland, where the parameter of y2 exceeds 0.15.In the majority of the analysed countries, producers specialized in pork can also be characterized by quite a large share of plant production.The share is higher than 30 % in 15 countries, namely in Austria, the Czech Republic, Germany, Denmark, Estonia, Finland, Hungary, Italy, Lithuania, Latvia, Poland, Portugal, Sweden and Slovenia.On the other hand, the share of plant and other animal production in total output is around 10% in the United Kingdom (UK) and the Netherlands, pointing to the high specialization of UK and Dutch farms.
Despite the pronounced differences in technology among the countries, some common patterns can be found.The elasticities for material inputs (specific and other material) with respect to outputs have the highest values, and the elasticities for capital are the lowest.Estonia is an exception.In Estonia, land has the highest elasticity (0.60), and the sum of material elasticities (specific and other) is the lowest (0.47) of all countries.That is, the impact of material inputs on production is the greatest among the analysed inputs.The same results can be found in Rizov et al. ( 2013), who analysed European old member states.Only in Estonia does an increase in agricultural land imply a higher increase in production than in material.The elasticity of capital is the highest in Italy (0.17).A high value for capital elasticity can also be found in France (0.15), Finland (0.13), Portugal (0.12) and Austria (0.11).However, there are no similarities in the impacts of other inputs (land, labour) in these five countries.For example, labour elasticity in Portugal is the highest of the analysed countries (0.26), while labour elasticity in France (0.06) has the second-lowest value.For Italian farms, on the other hand, land has the highest impact (0.28).
Table 2 provides estimates of the returns of scale.No indication of economies of scale (the sum of the elasticities is about one) was found for the average farm in the Czech Republic, Finland, France, Hungary, Lithuania and Slovenia.Bulgaria, Latvia and Romania are characterized by decreasing returns to scale.The other pork producers experienced increasing returns to scale.Thus scale efficiency has a large impact on productivity change in most member states.
Moreover, we can observe differences between OMS and NMS.Whereas OMS are characterized by increasing returns to scale (on average 1.08), the estimates indicate either constant or decreasing returns to scale for most NMS.According to Coelli et al. ( 2005), constant returns to scale are equivalent to the most productive scale size.That is, from a static point of view the farmers in OMS have a less-than-optimal size, and farmers in NMS are characterized by either an optimal or greater-thanoptimal size.However, significant differences among farmers in most member states can be found.The highest standard deviations were calculated for Finland, Italy, Portugal, Bulgaria, Estonia, Lithuania and Slovenia.  1 -as is the case for Greece, for example.OMS -old member states (EU -15), NMS -new member states.Source: author's own calculations Finally, Table 3 provides estimates of total factor productivity (TFP) indices and the contribution of the scale efficiency component to TFP change.TFP indices are calculated from the estimate of a metafrontier multiple output distance function for the analysed EU countries.As far as the level of total factor productivity is concerned, we can find significant differences among countries.In particular, OMS have a higher productivity level on average (1.12) compared to NMS (0.95).These results can also be affected by the Common Agricultural Policy (CAP), as mentioned by Rizov et al. ( 2013).The study by the mentioned authors concludes that coupled payments had a negative impact on farm productivity; however, decoupled payments contributed positively to TFP growth.Mary (2013) divided CAP subsidies into selective (e.g. investment subsidies, environmental subsidies) and automatic (e.g.single area payment, setaside premium, LFA) and found that automatic subsidies had a negative impact on TFP in France.Selective subsidies had no significant impact on production.The contribution of scale efficiency to productivity change is also higher, on average, in OMS compared to NMS.Moreover, it holds true in the majority of cases that the effect of scale efficiency was positive in OMS and negative in NMS.Austria, Finland and Slovakia are exceptions.The contribution of scale efficiency in OMS and NMS to TFP was also evaluated by Błažejczyk-Majka and Kala (2015) on the basis of the Malmquist index.However, their scale efficiency indices for livestock farms, which were slightly higher in NMS than in OMS, were not statistically significant.
The highest positive contribution of scale efficiency to productivity change was estimated for Spain and Italy.On the other hand, the highest negative impact of scale efficiency was observed in Estonia, Latvia and Romania.

CONCLUSION
In this section we will concentrate on the questions raised in the introduction, namely those regarding the exploitation of economies of scale from a static as well as dynamic perspective.The results show significant differences in the exploitation of economies of scale among the EU member states.In particular, OMS are characterized by increasing returns to scale, indicating a production of less-than-optimal size, whereas most NMS have either an optimal or greater-than-optimal size from a static point of view.
The differences among OMS and NMS are pronounced from a dynamic perspective as well.Whereas OMS improved productivity through scale efficiency, the impact of scale efficiency on productivity change in new member states was rather negative.
The results suggest that the competitiveness of the EU pork production can be increased through the scale efficiency.However, as far as the institutional support is concerned, policy makers should produce country specific actions by the exploitation of economies of scale.
For example, Mathijs and Swinnen (2001) analyzed the scale efficiency of East German farms; Bielik and Rajčániová (2004) did the same for farms in the Slovak Republic; Karagiannis and Sarris (2005) measured the scale efficiency of Greek farms; Latruffe et al. (2005) did the same for Poland; Madau (2012) calculated the scale efficiency of Italian farms; and Cechura (2014) and Cechura et al. (2014) did the same for Czech farms.A comparison of the EU countries was done by Blažejcyk-Majka et al. (2011) and more recently by Blažejcyk-Majka and Kala (2015).Their results (based on DEA and Farm Accountancy Data Network (FADN) data on average farms in EU regions for the period

Table 1 :
Sample descriptive statistics

Table 2 :
Returns to scale in EU pork production

Table 3 :
Contribution of scale effect to TFP change Source: author's own calculations