Impact of the EU Enlargement on the Agricultural Income Paweł Kobus, PhD, email: pawel_kobus@sggw.pl. Department of Agricultural Economics and International Economic Relations Warsaw University of Life Sciences Abstract The paper presents the analysis of the impact of the EU enlargement in 004 on the agricultural net value added and the factors income in the EU Member States. The main aim of the study is to test the significance of difference in reaction to the enlargement in three distinct groups of members, namely the old fifteen Member States, the new ten Member States which accessed the EU on May 1, 004, and the two newest Member States, i.e., Romania and Bulgaria which accessed the EU on January 1, 007. A linear mixed model was applied for the purpose of description of different countries behaviour. Key words: EU enlargement, agricultural output, mixed linear model. Introduction On May 1, 004 Cyprus, the Czech Republic, Estonia, Hungary, Latvia, Lithuania, Malta, Poland, Slovakia and Slovenia joined the EU. On that day the ten new Member States joined the single market. As it had been anticipated, the flow of trade between the old and the new Members States has amplified (Analysis 00). The EU funds have become available for farmers in the new Members States which allowed a significant increase of investments in agriculture and a certain economic boom in the rural areas. There is a general consensus that the EU enlargement has a positive effect on the EU agriculture as a whole. Nevertheless, some adjustments of production and consumption have had to take place in several countries of the EU-5 and not all of them for the better. The aim of this paper is to study the impact of the EU enlargement on agriculture in three distinct groups of countries: the old fifteen Member States (EU-15), the new ten Member States (EU-10N) which accessed the EU on May 1, 004 and the two newest Member States, i.e., Romania and Bulgaria (EU-N), which accessed the EU on January 1, 007. In a paper by the author Kobus (008) it was shown that the agricultural income in terms of indicator A had increased significantly only in the EU-10N, while for all other Members States it had remained, on average, on the previous level. What remains unclear is whether it could be just an effect of bigger subsidies in the EU-10N after accession? In order to answer this question the influence of enlargement on two characteristics of agriculture, namely the net value added and the factors income is investigated. Data The data used in this analysis are available from the Eurostat. According to a Eurostat guide (Manual 000) the net value added can be obtained by subtracting the fixed capital consumption from gross value added. The net value added is valued at basic prices, where a basic price is the price receivable by a producer from a purchaser for a unit of good or service produced as output plus any subsidy receivable on that unit as a consequence of its production or sale, minus any tax payable on that unit as a consequence of its production or sale. Factors income is the net value added enlarged by the value of other production subsidies minus other production taxes. The values of net value added and of factors income were originally expressed in EUR million (from January 1, 1999) or million ECU (up to December 31, 1998) at current prices. In order to eliminate the influence of price changes the values were deflated with producer price indices. In the next step both variables were transformed from absolute values to indexes, and the year 000 was chosen as a base year. Due to a limited range of available data, this study is based on the data starting in 1998 and ending in Economic Science for Rural Development Nr. 19., 009 63
007. For the same reason Cyprus and Estonia have been excluded and the data from only twenty five countries have been analysed. Statistical model As it was mentioned in the previous section the analysed data consisted, for each variable, of values applying to ten years and twenty five countries. As a result there are ten observations for each Member State, considering both variables individually. For the analysis of impact of the EU enlargement in the year 004 two explanatory variables were created: AfterAccess {0, 1}, a variable describing if an observation comes from a year after 004 and Group {A, B, C}, a variable which takes value A for countries from the EU-15, B for countries from the EU-10N and C for Romania and Bulgaria. The model which could be applied for such data is presented below (1): (1) where is the value of the response variable for the jth of n i observations in the ith country, is the value of the explanatory variable AfterAccess for the jth observation in the ith country, is equal 1 if the value of the explanatory variable Group for the jth observation in the ith country is B and 0 otherwise, is equal 1 if the value of the explanatory variable Group for the jth observation in the ith country is C and 0 otherwise, υ i, are random variables fully explained in the next paragraph, are the regression coefficients, which are identical for all groups. The parameter represents the so-called reference level which in this case applies to a situation when a country belongs to the EU-15 and the observation comes from a year before 004. There are two random variables in the model (1). First of them is υ i and represents the random effect in ith country, the second one is which represents the random error of jth observation from ith country. It is assumed that both variables follow normal distribution, with the expected value equal 0 and variances σ ν and σ ε respectively. It is also assumed that υ i for different values of i is independent, the same applies to which are also independent for different values of i and j. Hence if one of the regressors has a random character the model (1) belongs to the linear mixed models family (Demidenko 004). As it was mentioned in the introduction the aim of this paper is to study the impact of the EU enlargement on agriculture in three groups of countries. Thus three models were compared in order to assess the impact. Model (1), already presented, assumes that effect of the enlargement is the same in all three groups and that group effect is the same before and after the enlargement. Model () allows differences in reaction to the enlargement by including interaction terms. Model (3) contains only a constant besides random variables, and is equivalent to a lack of impact of the EU enlargement and also to a lack of differences between groups. Model (1) can be treated as a special case of model () with restrictions on two parameters ( ), also model (3) can be treated as a special case of model (1) with restrictions on three () (3) The net value added or the factors income 64 Economic Science for Rural Development Nr. 19, 009
parameters ( ). This allows the application of likelihood-ratio test for testing if the additional parameters are equal 0. LRT = ( LLF 1 LLF ) (4) 0 where: LRT is a value of the test statistic, LLF 1 and LLF 0 are values of a likelihood function logarithm calculated for appropriate models. If the hypothesis is true the LRT statistics follow asymptotically the chi-square distribution with p1 p 0 degrees of freedom (DF) distribution, where p 1 and p 0 are numbers of parameters for respective models. The difference p1 p0 is equal to the number of restrictions on parameters. The likelihood-ratio test can be used for testing hypotheses on a whole group of parameters at once and can be considered as a substitute for an analysis of variance test, when its assumptions are not fulfilled. In the further part of the paper model (3) will be denoted as M A0, model (1) as M A1 and model () as M A. The calculations for all models were performed in R, an environment for statistical computing (R: A language 008) with a help of the lme4 package (Bates 007). Results For the assessment of the changes in factors income three models were compared: M A0 containing only a constant, M A1 containing main effect of factors and M A containing main effects and interaction of factors. In order to test the significance of the influence of variables added in consecutive models a likelihood-ratio test was used. P-values presented in Table 1 correspond to two hypotheses: (5) (6) While the hypothesis (5) says that both the effect of the enlargement and the effect of group membership are nonexistent; the hypothesis (6) says only that the effect of enlargement is the same in all three groups, and that group effect is the same before and after the enlargement. The p-value corresponding to the hypothesis (5) is equal.1 E-07 and the p-value corresponding to the hypothesis (6) is even lower, hence both hypotheses can be rejected on 0.05 significance level. It indicates that the effects of both factors are significant as well as is their interaction. This suggests that reactions of countries from different groups to the EU enlargement in 004 differ. Results of testing the influence of explanatory variables on factors income indices Table 1 Model Number of parameters LLF LRT Chi DF p-value M A0 3 10.385 M A1 6 7.74 33.778 3.1E-07 M A 8 76.955 99.36 <.0E-16 Estimates of the M A model parameters for factors income indices Table Factor Parameter Estimate Std. Error t-value Intercept (reference level) 0.9785 0.085 34.3 Group B Group C AfterAcces 1 Group B: AfterAcces 1 Group C: AfterAcces 1 0 0.0715 0.0484 1.48 0.1630 0.0831 1.96 3 1-0.01 0.081-0.78 4 0.448 0.0477 9.8-0.335 0.080-4.09 5 Economic Science for Rural Development Nr. 19., 009 65
Fitted estimates of the M A model coefficients and t-statistic values for the hypotheses of equality of an appropriate parameter to 0 are presented in Table. Due to a big number of observations (50) a 0.975 quintile of the standard normal, i.e., 1.96 distribution is used instead of t-statistic as critical value. The value of intercept shown in Table and defined as a reference level is an average of the factors income indexes for countries from Group A (EU-15) before the year 004. The value of estimate for Group B is the difference between averages for the EU-10N and the EU-15. The same logic can be applied for the rest of factor levels. Figure 1 illustrates the mentioned differences in the reactions of three distinct groups to the EU enlargement. It may be clearly seen that Group demonstrates a big leap in factors income while Group 1 stays at the same level and Group 3 shows a significant decrease. The lines drawn on the figure can be understood as an indication of a relative direction. This is a typical way for the presentation of interactions between categorical variables. The fact that there is no evidence of significant changes in factors income (measured by indices) after accession only for the EU-15 is worth mentioning. So one must conclude that the countries which experienced the effect of the EU enlargement in 004 where not only the countries which actually accessed the EU in that year but also Bulgaria and Romania. On the average for the EU-10N, it was an increase of 4% of factors income, which is a sum of -.1% due to the main effect of the EU enlargement common to all groups and 44.8% due to the specific effect of the EU enlargement in Group B, compared to the years before accession. For the EU-N there was a decrease of about 35%. Similar analyses were performed to test whether the impact of enlargement was also significant for the net value added. The p-value corresponding to the hypothesis (5) is equal 3.78E-09 and the p-value corresponding to the hypothesis (6) is 3.74E-10, hence both hypotheses can be rejected on 0.05 significance level. This suggests that the changes of the countries from different groups to the EU enlargement in 004 are different. The estimates the M A model parameters, presented in Table 4, show whether the differences for the net added value are similar to that of factors income. Group 1 Group Group 3 160% 150% 140% 130% 10% 110% 100% 90% 80% 70% 60% Before accesion After accesion Figure 1. Interactions between variables Group and AfterAccess for factors income indexes Results of testing the influence of explanatory variables on the net added value indices Table 3 Model Number of parameters LLF LRT Chi DF p-value M A0 3 5.063 M A1 6 6.14 4.1 3 3.78E-09 M A 8 47.964 43.679 3.74E-10 66 Economic Science for Rural Development Nr. 19, 009
Estimates of the M A model parameters for the net added value indices Table 4 Factor Parameter Estimate Std. Error t-value Intercept (reference level) 0.966 0.0345 7.91 Group Group 3 AfterAcces 1 Group : AfterAccess 1 Group 3: AfterAccess 1 0 0.1313 0.0585.45 0.153 0.1006 1.515 3 1-0.56 0.0313-7.1 4 0.360 0.0531 6.145-0.1659 0.091-1.8 5 Group 1 Group Group 3 130% 10% 110% 100% 90% 80% 70% 60% Before accesion After accesion Figure. Interactions between variables Group and AfterAccess for net added value indices The results presented in Table 4 for the net value added indices are surprisingly different from such estimates for the factors income (Table ). This time the increase for the EU-10N is only 10% while for factors income it was 4%. For the EU-N the decrease is 39% and is similar to the previous one. The result for the EU-15, where the decrease of the net value added is estimated at.5%, while for the factors income it was only about % is the most surprising. As was mentioned in the previous part of the paper the lines drawn on Figure 1 and Figure show only directions of changes, and Figure shows that the net value added in the EU-15 has dropped significantly. Conclusions The impact of the EU enlargement on income in agriculture measured by factors income was positive for the new Member States which accessed the EU in 004, neutral for the old EU-15 and negative for Romania and Bulgaria. The net added value, which is the only part of income generated directly by agricultural holdings, also increased in the new EU-10N, but not as much as the factors income, 10% and 4% respectively. In the old EU-15, and Romania and Bulgaria, the net added value decreased significantly. Those differences in reaction to the EU enlargement in 004 for factors income and the net added value can be explained only by changes in the level of subsidies, which increased significantly and shrouded the decrease of income generated directly in agricultural holdings in the old Members States. In the new 10 Members States the income generated in agricultural holdings increased but not so much as could be thought after the analysis of factors income or indicator A (Kobus 008). Economic Science for Rural Development Nr. 19., 009 67
Bibliography 1.. 3. 4. 5. Analysis of the Impact on Agricultural Markets and Incomes of the EU Enlargement to the CEECs. [00]. Report prepared by the Directorate General for Agriculture of the European Commission. Bates D. [007]: lme4: Linear mixed-effects models using S4 classes. R package version 0.99875-9. Demidenko E. [004]: Mixed Models, Theory and Applications. Wiley, Nowy Jork. Eurostat, Statistical Office of the European Communities. [accessed 008]. http:// ec.europa. eu/eurostat. Kobus P. [008]: The impact of the EU enlargement on the agricultural output and 6. 7. 8. income in the member states. Scientific Journal Warsaw University of Life Sciences SGGW. Problems of World Agriculture v. 3. Manual on the Economic Accounts for Agriculture and Forestry EAA/EAF 97 (REV. 1.1). [000]. R: A language and environment for statistical computing. [008]. R Development Core Team. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, [available at] URL http://www.r-project.org. Regulation (Ec) No 138/004 of the European Parliament and of the Council of 5 December 003 on the economic accounts for agriculture in the Community. [004]. 68 Economic Science for Rural Development Nr. 19, 009