WPS8638 Policy Research Working Paper 8638 Poverty, Inequality, and Agriculture in the EU João Pedro Azevedo Rogier J. E. van den Brink Paul Corral Montserrat Ávila Hongxi Zhao Mohammad-Hadi Mostafavi Poverty and Equity Global Practice November 2018 Policy Research Working Paper 8638 Abstract Boosting convergence and shared prosperity in the European then evaluates at the regional level where the Common Union achieved renewed urgency after the global financial Agricultural Program funding tends to go, poverty-wise, crisis of 2008. This paper assesses the role of agriculture within each country. This approach enables making more and the Common Agricultural Program in achieving this. informed policy recommendations on the current state of The paper sheds light on the relationship between poverty the Common Agricultural Program funding, as well as eval- and agriculture as part of the process of structural trans- uating the role of agriculture as a driver of shared prosperity. formation. It positions each member country on the path The analysis performed throughout the paper uses a combi- toward a successful structural transformation. The paper nation of data sources at several spatial levels. This paper is a product of the Poverty and Equity Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/research. The authors may be contacted at jazevedo@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Poverty, Inequality, and Agriculture in the EU1  João Pedro Azevedo, Rogier J. E. van den Brink, Paul Corral, Montserrat Ávila, Hongxi Zhao, and  Mohammad‐Hadi Mostafavi  JEL‐codes: I32; N54; R12; R14  Keywords: Europe, Agriculture, Regional Development, Spatial Poverty and Spatial Inequality  1 The authors would like to thank the comments and suggestions from Jo Swinnen, Maria Garrone and Dorien Emmers (University of Leuven); Alessandro Olper (University of Milan); Alan Matthews (Department of Economics, Trinity College Dublin); Attila Jambor (Corvinus University, Budapest), Anastassios Haniotis, Director for Strategy, Simplification and Policy Analysis of DG AGRI (European Commission). The authors would also like to acknowledge the guidance and support from Arup Banerji (Regional Director, European Union, The World Bank), Lalita Moorty, Luis-Felipe Lopez-Calva and Julian Lampietti (Practice Managers, The World Bank). The usual disclaimer applies.   1. Introduction2    Shared prosperity is still a challenge in the European Union    After the global financial crisis of 2008, economic growth seems to be back on track in the European  Union.  Nevertheless,  disparities  across  EU  member  states  in  income,  growth,  and  the  speed  of  recovery, among other economic indicators, persist and remain to be solved. In this regard, it is also  worth mentioning that poverty rates for some EU countries are still higher than pre‐crisis levels. It  is clear then, that convergence and shared prosperity in the EU have room left for improvement.  Policies  that  promote  shared  prosperity,  by  ensuring  that  growth  reaches  everyone,  should  be  implemented. The focus of the present paper is agriculture and its role in ensuring shared prosperity  and fostering inclusive growth. We assess agriculture also from the role it plays in the structural  transformation of a country. Furthermore, we explore how the CAP program, being a policy targeted  to agriculture, has impacted shared prosperity and inclusive growth. In this fashion, we seek to point  out possible room for improvement in the CAP program and its current implementation, as well as  provide some general recommendations.    This working paper is organized around two main questions related to non‐intentional impacts of  the  CAP  in  respect  to  poverty  and  inequality  at  the  subnational  level.  Although  the  original  objectives of the CAP program were not necessarily aligned towards poverty alleviation, this paper  aims to first investigate whether it played a role in the registered reduction of monetary poverty  during the last decade. The second guiding question for this working paper consists on documenting  the relationship between agricultural activity and the CAP in respect to monetary poverty, with a  particular focus on the observed heterogeneity across EU member states. The answers to these two  guiding questions based on the latest and most granular analysis of the past ten years of the CAP,  can  inform  if  and  where  agriculture  and  the  CAP  can  be  one  of  the  important  drivers  of  social  inclusion and territorial cohesion.     A main motivating question in this analysis is whether, and how, the CAP may complement other  policies,  or  foster  territorial  cohesion  on  its  own.  A  clear  objective  of  the  EU,  as  stated  by  the  European  Commission,  is  to  “strengthen  economic  and  social  cohesion  by  reducing  disparities  between  regions  in  the  EU”  (European  Parliament,  n.d.).  In  addition  to  economic  and  social  cohesion, territorial cohesion was later included as a further objective. In a similar line as Crescenzi  and Giua (Crescenzi & Giua, 2016), an important motivating question is whether sectorial policies  like the CAP can contribute to or complement other policies’ objectives, specifically social cohesion  in the EU. In the particular case of the CAP, we want to analyze if this program further complements  other  existing  policies’  objectives  by  better  channeling  resources  to  socio‐economically  deprived  areas.  This  could  potentially  aid  in  the  Cohesion  Policy  of  the  EU,  as  it  contributes  to  reducing  disparities between regions, based on poverty rates, across the EU.     The  data  used  for  the  purpose  of  the  various  analyses  in  this  working  paper  came  from  several  sources; including the EU‐SILC survey from years 2003 to 2014 at the NUTS 1 and NUTS 2 levels, the  2 This work was produced as a background paper to the EU Regular Economic Report : Thinking CAP ‐ Supporting Agricultural Jobs and Incomes in the EU. (Brink, Kordik, and Azevedo, 2018). 2 EU‐Poverty Map at the NUTS 3 level, the CAP administrative records at the NUTS 3 levels, and the  Farm Structure Survey for several years including 2010 and 2011. This paper is structured as follows.  In  Section  2  we  start  by  briefly  motivating  the  need  for  additional  policies  that  foster  inclusive  growth,  as  suggested  by  the  state  of  inequality  and  poverty  in  the  post‐crisis  period.  Section  3  introduces  the  main  framework,  and  the  results  from  the  analysis  of  the  relationship  between  poverty  and  agriculture.  Section  4  then  briefly  introduces  the  CAP  and  proceeds  to  analyze  the  relationship between poverty and the CAP. Section 5 wraps up the main results, by presenting an  integrative analysis taking results from the previous sections, and finally Section 6 concludes.     2. Motivation    The current state of inequality and poverty in the European Union    After the global financial crisis of 2008, some of the economic indicators across the European Union  have been under recovery, however others still lag behind, one of these being inequality. Thus, the  current  state  of  inequality  and  poverty  in  the  EU  points  to  the  need  for  inclusive  policies  that  promote shared prosperity. In this section we describe the current state of these indicators in the  region. Convergence in agricultural income is also introduced and briefly discussed as an important  channel for overall income convergence across EU member states.    Inequality has become an important topic in the policy discussion of the EU, especially since the  Great Recession. Even though inequality in the EU member states is low compared to other parts of  the  developed  world  (OECD,  2017),  inequality  in  the  region  has become  a  topic  of  concern.  The  recent  member  state  expansion  towards  countries  with  lower  levels  of  average  income  has  contributed to an increase of inequality across the EU. In order to further explore this, we perform  an analysis by treating the EU as a single country, thus pooling the income of all member countries  together and ranking them along the same distribution. The resulting Gini coefficient is higher than  the  coefficient  associated  to  any  single  EU  member  state.  This  is  a  high  inequality  level  by  international standards.     Figure 1. Inequality Across the EU 0.4 0.35 Gini by Member States and Pooled EU 0.3 0.25 0.2 0.15 0.1 0.05 0   3 Source: EUROSTAT, WB staff calculations.      Although  poverty  is  multidimensional,  we  focus  on  a  single  dimension  with  the  purpose  of  maximizing comparability among the various data sources used. Thus, in the context of this work  we focus exclusively on the monetary dimension of poverty, using one main measure, namely, the  2011‐anchored AROP (at risk of poverty). This measure uses a relative poverty line for each member,  but keeping it constant for all the years within the analysis. A country’s poverty line is defined as  60% of its equivalized median income anchored on the 2011 value. Figure 2 shows the trends from  2004  to  2014  for  GDP  per  capita  and  anchored  monetary  poverty  (using  an  anchored  relative  poverty  line  for  each  member  state).  Since  the  survey  coverage  changes  over  time  due  to  the  expansion of the EU membership, we compute separate lines for each cohort of member states in  terms  of  comparable  data  availability.  The  figure  shows  that  although  GDP  per  capita  is  already  above its pre‐crisis level, the relative anchored poverty level is at higher or at the same pre‐crisis  level, suggesting that growth during the recovery has not been inclusive. The case for the Southern  EU member states is particularly alarming.    Figure 2. Although GDP per capita has recovered, anchored relative poverty rates are still higher.   Source: EUROSTAT, WB staff calculations.    Average Anchored Relative Poverty Rate   GDP per inhabitant, PPS 0.4 35 Thousands 0.35 30 0.3 25 0.25 20 0.2 15 0.15 10 0.1 0.05 5 0 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Central EU Northern EU Central EU Northern EU Southern EU Western EU Southern EU Western EU   Furthermore,  absolute  poverty  levels  remain  high  across  the  EU.  In  order  to  compare  absolute  poverty levels between countries, we define the median of the absolute national poverty lines of all  EU  Member  States  as  the  absolute  poverty  line  to  use  across  countries.  This  gives  an  absolute  poverty line of US$21.70 per day (in PPP). Using this measure, poverty remains high across the EU,  and  further  stresses  the  large  disparities  across  countries.  Figure  3  below  compares  the  poverty  rates that result from using absolute measure and relative measures.     Figure 3. Poverty rates between member states are extremely different using an absolute poverty line, compared to a relative measure.  4 1 Percentage of population living under 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 Median EU at risk of poverty line At risk of poverty line   Source: EUROSTAT, WB staff calculations.      The role of agriculture in income convergence    Although the speed of convergence remains low, member states are catching up with each other in  terms  of  income,  and  at  a  faster  rate  for  agricultural  income.  In  recent  years  there  has  been  a  reduction in the dispersion of mean incomes between EU member states, or what is referred to as  beta convergence in the literature. This means that member states have experienced a convergence  in their income level, especially in agricultural income. In particular, agricultural income growth is  converging  faster  than  non‐agricultural  income  growth.  This  may  also  indicate  a  decrease  in  the  agricultural income gap. However, results show that agricultural income is catching up faster with  non‐agricultural income in old member states (OMS), compared to the pace for new member states  (NMS). This result further highlights the importance of addressing how other policies can contribute  towards  more  inclusive  growth,  in  particular  the  CAP,  which  targets  agriculture  and  may  help  in  reducing  the  difference  in  convergence  rates  between  member  states.  The  potential  of  the  CAP  program in aiding inclusive growth depends on the state of structural transformation where each  country finds itself.    Table 1. Speed of convergence for different incomes between OMS and NMS OLS LSDV EU27 OMS NMS EU27 OMS NMS Speed of Convergence, β 0.04% 0.00% 0.05% 0.21% 0.21% 0.24% Mean total household income Half-life of convergence 1863 175913 1340 334 338 290 Speed of Convergence, β 0.13% 0.13% 0.08% 0.72% 0.76% 0.65% Mean agricultural income Half-life of convergence 550 525 903 97 91 107 Notes: (1) Speed of Convergence is calculated using the coefficients of respective variables of interests; (2) Half-life of convergence is calculated as 0.6931/speed of convergence; (3) OLS columns report results using OLS regressions, while LSDV columns results using fixed effect models; (4) OMS are 14 old member states, while NMS are 13 new member states that joined the EU after 2004; (5) Data Source: EU-SILC, Eurostat (2005-2014)   5 3. Agriculture and Poverty    A Structural Transformation Approach to agriculture and poverty    Structural transformation may be broadly defined as the transition of an economy from a strong  reliance in labor intensive and low‐productive sectors to more skill‐intensive and high‐productive  sectors (UN Habitat, 2016). This transition usually occurs as labor and other economic resources  move  away  from  the  traditionally  labor‐intensive  agricultural  sector  to  modern  sectors  such  as  manufacturing and services, which are characterized by higher skills and productivity. Concomitant  to  such  transition  is  an  increase  in  productivity  and  income.  There  are  several  economic  characteristics  of  the  agricultural  sector  within  a  country,  which  signal  an  ongoing  structural  transformation. For instance; a declining share of the sector’s contribution to GDP, migration from  rural to urban areas, which at the same time results in a decrease of the sector’s share of overall  employment,  an  increase  in  agricultural  labor  productivity,  and  eventually  a  decline  in  poverty,  among others.    Figure 4. We assess the key relationship between poverty, agriculture, and the CAP, from a structural transformation approach   Agriculture Structural Transformation Poverty CAP   For  this  work,  we  operationalize  the  process  of  structural  transformation  as  described  in  what  follows. As the transformation takes place, the agricultural sector gains in competitiveness, and its  productivity  increases.  Agriculture  then  represents  a  source  of  growth  and  jobs  for  the  regions  where it is a predominant economic activity. Thus, it starts by spurring growth in such regions and  in this way, contributes to a decrease in the associated poverty. As agricultural labor productivity  increases, agricultural income increases to a point where poverty reduction is first observed in the  immediate  areas  where  agriculture  predominates.  As  incomes  continue  to  increase,  the  effect  extends to whole rural areas in such a way that rural poverty is reduced. Furthermore, past this  point  of  the  process,  agriculture  and  poverty  start  to  appear  negatively  associated.  Structural  transformation thus hints at the role that agriculture as a sector plays in promoting inclusive growth  and shared prosperity, as it represents the first milestone of a successful transformation. It is along  this line that we continue our analysis in what follows.  6 Identifying successful and incomplete transformers using the poverty rate and agricultural indicators    Following this approach, we perform a first analysis to explore the association between poverty and  several agricultural indicators, which assess the extent of agricultural activity within a region. In this  fashion we seek to identify where a country is currently located on the path towards a successful  structural  transformation.  The  stylized  story  behind  this  is  that,  as  mentioned  earlier,  low  productivity in agriculture translates into high poverty in the areas where agriculture prevails. As  the transformation moves forward, agriculture becomes more productive and incomes expand, thus  decreasing poverty in agricultural regions. It is important to identify regions in which agricultural  activity remains closely associated to poverty, as this suggests that they remain in an early stage of  a  structural  transformation  and  may  still  have  untapped  opportunities  to  accelerate  their  development process in the near future. For this purpose, we create six indicators that capture the  intensity  of  agricultural  activity  within  a  region;  share  of  agricultural  area,  average  agricultural  output per hectare, average labor unit per hectare, average labor unit per holding, average holding  size, and agriculture share of employment. The first analysis performed consists of assessing, for  each country, how each of these indicators is correlated with poverty. Following the stylized story  on structural transformation, a negative association between an agricultural indicator and poverty  signals  a  successful  structural  transformation,  while  a  positive  association  signals  room  for  improvement within the transformation path.     We  consider  two  measures  of  area  poverty:  poverty  rate  and  the  share  of  a  country’s  poor  population. We start with the spatial distribution of poverty, measured by the regional poverty rate,  within each member state. This indicator identifies the regions in which poverty tends to happen.  The  results  are  summarized  in  the  following  table,  where  the  sign  captures  the  direction  of  a  statistically  significant  association  found  between  the  poverty  rate  and  the  specific  indicator  referred to in each column, while controlling for a number of observation factors such as population  and GDP. A zero indicates that no significant association was found. Thus, a positive sign suggests  that agricultural activities, as measured by the indicator in place, tend to take place in poorer regions  within  a  country.  In  a  similar  fashion,  a  negative  sign  suggests  that  such  activities  tend  to  concentrate in non‐poor regions.    Table 2. Association between poverty rate and different agricultural indicators  Average Average labor Average Average Agriculture Agriculture holding Average labor unit per agricultural output per share of share of size unit per holding output per labor unit Country area employment (hectare) hectare (AWU) (AWU) hectare (Euro) (Euro/AWU) Croatia + + + + + - + Spain + + + + + + + Bulgaria + + + + + + + Portugal + + + + + + + Slovenia + + - + + + + Latvia + - + + + + + Greece + + - + + + - Romania + - + + - + - Malta + + 0 + + + + Italy + + - + - + + Sweden + + - - + - + United Kingdom - + - - - + + 7 Estonia - + - - + - - Germany + - - - + + + France - - - - - - - Ireland - + - + - - - Slovak Republic - - - - - - - Austria - - - - - - - Finland - + - + + + - Poland - - - - - - - Belgium - - - - - + + Hungary - - - - - - -    Table  2  shows  the  heterogeneity  across  the  EU  regarding  the  stage  of  structural  transformation  where its member states find themselves, suggested by the association with different sign patterns  for the various indicators. The successful transformers show a negative correlation between poverty  and  agricultural  indicators,  consistent  with  the  fact  that  at  this  phase  of  the  transformation  agriculture is no longer linked to poverty. Such is the case for Austria, France, Hungary, Poland, and  the  Slovak  Republic.  On  the  other  hand,  incomplete  transformers  show  a  consistent  positive  correlation  between  agricultural  activity  and  poverty,  as  agriculture  is  still  predominant  in  poor  regions.  Spain,  Bulgaria,  and  Portugal  are  among  the  countries  at  an  early  phase  of  the  transformation.    Identifying  successful  and  incomplete  transformers  using  the  share  of  the  country’s  poor  and  agricultural indicators    Following the analysis, we continue by assessing the correlations between poverty and agricultural  activity at the regional level, but now using as a measure of poverty the share of a country’s poor in  each region. The share of a country’s poor population concentrated in a particular region contrasts  with the poverty rate, as the former one is informative in terms of where the poor population tends  to concentrate, rather than where poverty tends to happen. It is important to distinguish between  the two poverty measures used, since the regions with high poverty do not necessarily contain a  higher share of the country’s poor population. This exercise sheds light on whether agriculture takes  place in areas with a high proportion of the total poor population within each EU member state. The  results are summarized in Table 3, where once again a positive sign within a specific agricultural  indicator denotes that agricultural activity, as measured by such indicator, takes place in regions  where poor people tend to concentrate. Analogously, a negative sign for an indicator underpins that  agricultural activity takes place in regions with a low concentration of the country’s poor people.    Table 3. Association between share of country’s poor and different agricultural indicators  Share of the Share of the Average Average labor Average labor Average Average country country holding unit per unit per agricultural output per agriculture agriculture size hectare holding output per labor unit Country area employment (hectare) (AWU) (AWU) hectare (Euro) (Euro/AWU) Latvia + + - + - - + Estonia + + + + - + + Ireland + + + - - + + Denmark + + + + - + + Slovak Republic + - + - - + + 8 Croatia + + + + - + + Portugal + - - + - + + Austria - - - + - + + Bulgaria - - - + - + + Hungary - - - + - + + Sweden - - - - - + + Finland - - + - - + + Romania - - - + - + + Greece - - - + - + + Belgium - - - + - + + Germany - - - + - + + Poland - - - + - + + Slovenia - - + + - + + Netherlands - - - + - + + Italy - - - + - + + France - - - + - + + Malta - - - - + - -   From Table 3 it is clear that there is heterogeneity across the EU regarding the relationship between  the share of poor and agricultural activities. Similarly as before, in this case successful transformers  can be identified as those where agricultural activities are negatively associated with the country’s  share  of  poor.  Such  is  the  case  of  Malta,  Sweden,  and  the  Netherlands,  among  other  countries.  Incomplete transformers still show a positive association between the share of poor and agriculture,  suggesting  the  prevalence  of  agricultural  activities  in  the  regions  where  the  poor  tend  to  concentrate  the  most.  In  this  case  we  find  countries  like  Croatia,  Estonia,  Ireland,  and  Portugal,  among others.    The analysis made so far will be complemented in what follows, by assessing the regions, in terms  of poverty, where the CAP funds tend to go. In this fashion we seek to better evaluate potential  improvement areas for the CAP funds within each member state.     4. Poverty, Inequality and the CAP    Assessing the CAP: A brief introduction    The Common Agricultural Policy was created in 1962 (European Commission, 2017) and thus stands  as one of the oldest policies of the EU. According to the European Commission, the main objectives  of the CAP today are “to provide a stable, sustainably produced supply of safe food at affordable  prices for Europeans, while also ensuring a decent standard of living for farmers and agricultural  workers” (European Commission, 2017). Broadly speaking, the Common Agricultural Policy has two  main components: pillar 1 and pillar 2. Pillar 1 consists of direct payments and market measures.  Under  this  pillar  farmers  can  receive  coupled  direct  payments,  which  are  conditional  to  the  production  of  a  particular  crop  or  livestock  species.  This  pillar  also  entitles  farmers  to  receive  decoupled payments, which do not depend on output, but on the area of the agricultural land used.  On  the  other  hand,  pillar  2  focuses  on  funding  rural  development  projects.  These  funds  support  investment in development projects taken on by farmers or rural businesses.  9   Figure 5. Levels of CAP funds received by different member states are drastically different. Total CAP Payments (2008 - 2013) 8E+10 Purchasing Power Standard 7E+10 6E+10 5E+10 4E+10 3E+10 2E+10 1E+10 0   Source: DG AGRI (2017) Clearance Audit Trail System (CATS) database provided by the European Commission    Figure 5 above shows how the allocation of total CAP payments differs across the EU member states.  The  EU  determines  how  much  CAP  funds  each  of  the  member  states  receives.  Nevertheless,  member states have certain flexibility in allocating the CAP funds between the program’s pillars.  This creates heterogeneity as to how the funds are spent, between both pillars, within countries, as  showed in Figure 6 below.  Figure 6. There is clear heterogeneity of CAP funds allocation choices made by different countries. Composition share of CAP payment by country (2008-2013) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% LFA Agrienvironment Investment aid Other Pillar 2 payments Decoupled payments Coupled payments   Source: DG AGRI (2017) Clearance Audit Trail System (CATS) database provided by the European Commission      Allocation of CAP funds based on regional characteristics    To explore the characteristics of regions where the CAP funds tend to reach, we create four different  clusters of regions at the NUTS1/NUTS2 levels based on the average holding size and the number of  employees  per  holding.  For  each  of  these  variables  we  create  categories  to  group  the  existing  regions. Table 4 below simplifies the clusters by characteristics, in which regions are grouped. Table  10 5 provides summary statistics for each of the four clusters. The holding size average for all regions  is 36.1, while the average number of workers per holding stands at 1.4. Clusters 1 and 2 contain  regions with mostly small holdings, whereas clusters 3 and 4 contain regions with a higher number  of employees per holding. Most of the regions belong to clusters 1 and 2.  Table 4. Clusters of regions based on average holding size and employees per holding Average holding size Small Average Large Low Cluster 1 Cluster 2 Employees per holding High Cluster 3 Cluster 4    Table 5. Summary Statistics for clusters Total number Mean holding Mean employee per of Number of Cluster Qualitative Interpretation size holding holding regions s in 2013 Cluster 16.43 1.01 9697200 60 1 Small Holding, Low Employment Cluster Average Holding, Low 55.97 1.35 638650 26 2 Employment Cluster Average holding; High 97.33 2.47 146360 14 3 employment Cluster 146.49 3.65 47380 6 4 Large holding; High employment   Figure 7(a) also shows that clusters 1 and 2 receive over 90% of the CAP funding, while cluster 4 is  the one that receives the least. Figure 7(b) contains the CAP composition by clusters. It shows that  for all clusters, most of the CAP funding is allocated to decoupled payments, followed by coupled  payments for clusters 1, 2, and 3. Thus the majority of CAP funding is allocated to pillar 1.    Figure 7. Regions in different clusters are receiving drastically different levels of CAP fund and have significant different allocation choice for their CAP fund. (a). Share of each CAP type received by each cluster 11 Cluster share of CAP Payments (Cluster created by Holding size and Labor unit per holding) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% LFA Agrienvironment Investment aids Other Pillar 2 Coupled Decoupled payments payments payments Cluster 1 Cluster 2 Cluster 3 Cluster 4 (b). CAP composition by clusters CAP Composition Payment by Clusters (Cluster created by Holding size and Labor unit per holding) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Cluster 1 Cluster 2 Cluster 3 Cluster 4 LFA Agrienvironment Investment aids Other Pillar 2 payments Coupled payments Decoupled payments Are the CAP funds reaching the poor regions within the EU?    Heterogeneity  of  the  CAP  funds  received  by  each  member  state  and  on  how  they  are  allocated  between the program’s pillars raises the question of whether heterogeneity persists, regarding the  characteristics of the areas where the CAP funds reach within each country. In particular, following  our  structural  transformation  approach,  we  are  interested  in  assessing  whether  there  is  any  relationship between poverty and CAP funds. We start by investigating where the CAP funds are  being allocated, by analyzing the relationship between the CAP funds and the spatial poverty rate.    Table 6. CAP payments are allocated to poorer regions 12 NUTS 3 CAP regressions (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) LABELS Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Payment (Total) 1.95e-09** Payment per capita (Total) 0.000472** Payment (pillar1) 1.95e-09** Payment per capita (pillar1) 0.000433* Payment share (pillar1) -0.00773 Payment (pillar2) 8.26e-09** Payment per capita (pillar2) 0.00325** Payment share (pillar2) 0.00773 Payment (pillar1 coupled) 1.66e-09 Payment per capita (pillar1 coupled) 0.00108 Payment share (pillar1 coupled) 0.000739 Payment (pillar1 decoupled) 2.56e-09** Payment per capita (pillar1 decoupled) 0.000541* Payment share (pillar1 decoupled) 0.00195 GDP per capita (PPS) -0.000154** -0.000151** -0.000155** -0.000153** -0.000158** -0.000153** -0.000150** -0.000158** -0.000158** -0.000157** -0.000158** -0.000154** -0.000152** -0.000158** Population density 0.113** 0.111** 0.112** 0.109** 0.0990** 0.106** 0.113** 0.0990** 0.102** 0.103** 0.100** 0.114** 0.110** 0.103** Country fixed effects Y Y Y Y Y Y Y Y Y Y Y Y Y Y Constant 17.54** 17.23** 17.79** 17.65** 18.60** 17.14** 15.68** 17.83** 18.15** 18.01** 18.19** 17.72** 17.60** 18.10** Observations 1,181 1,181 1,184 1,184 1,181 1,181 1,181 1,181 1,181 1,181 1,181 1,184 1,184 1,181 Adj R-Squared 0.368 0.364 0.363 0.360 0.359 0.372 0.371 0.359 0.359 0.360 0.359 0.365 0.360 0.359 Note: (1) Data source: EU 2011 Poverty Map (DG-REGIO and World Bank), 2010 Farm Structure Survey (Eurostat), and EU National Statistic Institutes (Eurostat); (2) Symbols: ** p<0.01, * p<0.05, + p<0.1; (3) Missing observations in CAP data are treated as zero; (4) Luxembourg and Cyprus Republic are not being analyzed here;(5) CAP data is missing for 21 regions in Croatia and 2 regions in Spain (with 1 more region missing for Pillar 2 data); (6) Standard errors are clustered at country level   Table 6 above indicates that, overall, the CAP funds seem to be reaching poorer areas within the EU.  Starting  with  total  CAP  payments,  a  positive  and  significant  relationship  with  poverty  rate  is  observed. Further disaggregating the total payments into the two CAP pillars supports this result, as  both  pillars,  when  individually  analyzed,  remain  significantly  associated  to  poverty  rates.  When  analyzing pillar 1 by its individual components, the decoupled payments remain significantly and  positively  associated  with  poverty  rate.  These  results  remain  when  analyzing  each  of  these  components  on  a  per  capita  level.  Therefore,  CAP  funds  do  tend  to  reach  regions  where  higher  poverty rates prevail. Nevertheless, it is important to keep in mind that CAP payments reaching poor  areas do not imply they are necessarily reaching the poorest households within those areas.    We continue our analysis by assessing the relationship between  CAP payments and poverty, but  now turning to the share of the countries’ poor as the measure for poverty. Table 7 below presents  the results obtained, which are similar to the ones previously presented using the poverty rate. Total  CAP  payments  are  significantly  and  positively  associated  with  the  share  of  the  country’s  poor.  Similarly, for the specifications including each of the CAP’s pillars, we observe the same significant  and positive association. In this case, both components of pillar 1, coupled and decoupled payments,  seem to be reaching areas where there is a high share of the country’s poor population.     Table 7. CAP payments (1) (2) (3) (4) (5) (6) (8) (10) (12) (14) (16) (18) Share of Share of Share of Share of Share of Share of Share of Share of Share of Share of Share of Share of country country country country country country country country country country country country LABELS poor poor poor poor poor poor poor poor poor poor poor poor Total CAP payments 1.46e-09** 1.45e-09** Payments for pillar1 1.39e-09** 1.83e-09** Payments for pillar2 6.73e-09* 6.11e-09* Payments for pillar1 coupled 4.46e-09* Payments for pillar1 decoupled 2.04e-09** Payments for investment aids budgets 1.08e-08** Payments for LFA budgets 2.22e-08+ Payments for Agri-environmental budgets 2.00e-08** Payments for pillar2 other 1.82e-08** Population density(Inhabitants per hectare); 0.0571* 0.0511* 0.0527* 0.0427* 0.0525* 0.0534+ 0.0460* 0.0483* 0.0416* Gross Domestic Product(PPS per inhabitant) 6.98e-06 9.42e-06 7.18e-06 6.28e-06 9.64e-06 7.89e-06 6.21e-06 1.12e-05 -3.93e-06 Poverty line 0.000804** 0.000765** 0.000664** 0.000702** 0.000776** 0.000725** 0.000579** 0.000600** 0.000704** Number of zero-benefitiary budgets 0.0246+ 0.00257 0.0243* -0.0140 -0.479 0.0345+ 0.121+ 0.0524 0.212+ Constant 2.476** -11.33** 2.647** -9.422** 2.123** -9.474** -8.189** -9.595** -10.19** -7.082** -8.160** -8.186** Country fixed effects Y Y Y Y Y Y Y Y Y Y Y Y Observations 1,181 1,181 1,181 1,181 1,181 1,181 1,181 1,182 1,181 1,181 1,181 1,181 Adj R-squared 0.566 0.599 0.561 0.577 0.575 0.604 0.571 0.576 0.595 0.595 0.587 0.587 Note: (1) Data source: EU 2011 Poverty Map (DG-REGIO and World Bank), 2010 Farm Structure Survey (Eurostat), and EU National Statistic Institutes (Eurostat); (2) Symbols: ** p<0.01, * p<0.05, + p<0.1; (3) Missing observations in agricultural indicators are treated as zero;(4) Luxembourg and Cyprus Republic are not being analyzed here;(5) CAP payment data in Croatia, 5 regions in Estonia and 1 region in France (Seine-Saint-Denis) is missing;       13 Do the CAP payments reach poor regions within each country?    So far, the broad picture points to CAP payments, on average, reaching areas with higher poverty  rates and with a higher share of the countries’ poor population. Nevertheless, this need not be true  for  all  countries.  Therefore,  we  extend  the  analysis  to  assess  any  possible  heterogeneity  across  member  states.  For  this  purpose,  we  compute  the  relationship  between  the  poverty  rate  and  8  different  indicators  for  the  CAP  payments.  Such  indicators  include  the  total  payments,  then  disaggregate the total payments by pillars, and finally individually analyze the components within  each pillar. The results are presented in Table 8 below. This analysis provides spatial information as  to where CAP funds reach within each particular country, in terms of poor and non‐poor regions.     Table 8. Association between poverty rate and CAP payments  pillar1 pillar1 total pillar1 pillar2 coupled decoupled CAP CAP CAP CAP CAP LFA Agrienvironmental Investment Aids Country payments payments payments payments payments payments payments payments Croatia + + 0 + + 0 0 0 Spain + + + + + + + + Romania + + + + + + + + Bulgaria + + + + + + + + Portugal + + + + + + + + Slovenia + + + + + + + + Greece + + + + + + + + Italy + + + + + + + + Malta + 0 + 0 0 + + + Sweden + - - + + - + + Belgium - - - - - + + - Finland - - - - - - - + United Kingdom - - - - - - - - Latvia - + - + + - - - Ireland - - - - - - - - Germany - - - - - - - - Netherlands - - + - - + - - Austria - - - - - - - - France - - - - - - - - Poland - - - - - - - - Denmark - - - - - 0 - - Slovak Republic - - - - - - - - Estonia + - + - + - + +   The sign indicates the direction of a significant association found between the poverty rate and the  specific CAP payment indicator. A zero indicates that no significant association was found. However,  for the case of Croatia, a zero indicates missing data due to its more recent entry to the EU. Thus a  positive sign suggests that the particular CAP payment referred to by the indicator in place tends to  be allocated to poorer regions within a country. In a similar fashion, a negative sign suggests that  such payment reaches non‐poor regions. The heterogeneity of the spatial correlation between CAP  14 funds and poor regions across EU member states can be seen, as different countries are associated  with different patterns of signs for the indicators. Countries that allocate all CAP funding to high  poverty regions include Spain, Romania, Bulgaria, Portugal, Slovenia, Greece, and Italy. On the other  hand, countries for which CAP funds are negatively correlated with poverty rate include Hungary,  the Slovak Republic, Poland, France, Austria, the United Kingdom, Germany, and Ireland.    Table 9 presents the results for the analysis of the correlations between the CAP payment indicators  and the share of a country’s poor. Countries for which all the CAP payments reach areas where a  large share of the country’s poor population concentrates include Malta, Latvia, Ireland, Denmark,  and  Estonia.  On  the  other  hand,  countries  which  show  a  negative  correlation  between  CAP  payments and areas with a high concentration of the country’s poor include Spain, Italy, the United  Kingdom, Germany, France, Poland, and Hungary. The cases for countries like Spain and Ireland are  interesting,  since  both  countries  completely  switch  their  correlations  depending  on  the  poverty  indicator used. In the case of Spain, the country’s CAP payments reach regions with high poverty  rates,  but  are  negatively  correlated  with  regions  which  show  a  high  share  of  the  country’s  poor  population. Ireland shows the opposite case; its CAP payments consistently reach areas where the  poor concentrate, but are negatively correlated with regions that show high poverty rates.    Table 9. Association between share of poor and CAP payments  pillar1 pillar1 total pillar1 pillar2 coupled decoupled Investment CAP CAP CAP CAP CAP LFA Agrienvironmental Aids Country payments payments payments payments payments payments payments payments Croatia + + 0 + + 0 0 0 Spain - - - - - - - - Romania + + + - - - - - Bulgaria + + + + + - + - Portugal + + - + + - - - Slovenia + + + + + - + + Greece - - + - - - - - Italy - - - - - - - - Malta + + + + + + + + Sweden + + + + + - - + Belgium + + + - - - - + Finland + + + + + - - + United Kingdom - - - - - - - - Latvia + + + + + + + + Ireland + + + + + + + + Germany - - - - - - - - Netherlands + + + - - + - + Austria + - + - - - - - France - - - - - - - - Poland - - - - - - - - Denmark + + + + + + + + Slovak Republic + + - + + + - + Estonia + + + + + + + + 15 Hungary - - - - - - - -     After documenting that CAP funds tend to reach regions with high poverty rates and with a high  share  of  a  country’s  poor,  the  question  of  whether  CAP  has  actually  contributed  to  poverty  alleviation remains pending. It is worth highlighting that by answering this question, we are in no  way trying to evaluate the CAP’s overall performance since, as specified earlier in this paper, the  program’s  original  objective  is  not  related  directly  to  reducing  poverty  in  the  areas  where  it  is  allocated. Nevertheless, this question is important for further policy recommendations, and more  importantly,  to  assess  whether  CAP  is  an  instrument  that  supports  the  successful  structural  transformation of a country.     Poverty and inequality dynamics and the CAP    With this in mind, we use panel data to determine the impact, if any, that the CAP has had on poverty  rates. Despite heterogeneity in the allocation of the CAP across EU member states, on average a  positive impact of the CAP on poverty has been observed. Table 10 documents that total per capita  CAP payments are associated with a decrease in the poverty rate over time in the EU. The total per  capita payments are further disaggregated into the program pillars to explore potential differences  in each pillar’s contribution to poverty alleviation. In the individual analysis, the per capita payments  of both pillars remain significant and negative in their association with poverty growth. However,  when both pillars are jointly analyzed, it seems that pillar 2 is more significant in its contribution to  poverty reduction.     Table 10. Per capita CAP payments are linked to regions with higher poverty reduction. (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Total payments (per capita) -0.294* -0.185+ -0.244* Payments to pillar1 (per capita) -0.352* -0.173 -0.270+ -0.251 -0.0519 -0.171 Payments to pillar2 (per capita) -0.618* -0.517* -0.549* -0.393+ -0.476* -0.412+ Share of individuals in agricultural households 0.314+ 0.235 0.300* 0.216 0.344* 0.251+ 0.339* 0.246 Share of inhabitants with secondary education 0.00214 0.126 0.0166 0.137+ 0.0148 0.175+ 0.00786 0.135+ Share of inhabitants with tertiary education -0.137 -0.104 -0.131 -0.0995 -0.130 -0.0838 -0.133 -0.100 Unemployment rate 0.420* 0.429* 0.470* 0.454** GDP per inhabitant -2.791 -4.405 -2.760 -4.288 -3.186 -5.179 -3.090 -4.649 Dependency ratio -0.0620 -0.152+ -0.0443 -0.136 -0.0601 -0.152* -0.0650 -0.158+ Population density 0.00495* 0.00307 0.00617** 0.00433* 0.00445* 0.00302+ 0.00428* 0.00260 R-Squared: Within 0.387 0.497 0.433 0.367 0.480 0.415 0.359 0.508 0.425 0.388 0.510 0.437 Between 0.0586 0.222 0.0453 0.0416 0.213 0.0435 0.0537 0.330 0.0964 0.0663 0.312 0.0588 Overall 0.00414 0.283 0.0866 0.00302 0.266 0.0789 0.0427 0.375 0.133 0.00611 0.363 0.103 Observations 959 959 959 959 959 959 959 959 959 959 959 959 Notes: (1) Panel Fixed Effects; All regressions include year fixed effects; (2) Anchored Relative Poverty line (60% of median at 2011); (3) Data Source: Farm Structure Survey, Eurostat; EU-SILC, Eurostat ; (4) Symbols: ** P<0.01, * P<0.05, +P<0.1 In  a  similar  way  as  with  poverty,  our  analysis  also  finds  a  significant  impact  of  the  CAP  on  the  dynamics of inequality within regions in the EU. For this purpose, we first draw on the Gini index to  measure  inequality.  Starting  with  per  capita  total  payments,  we  find  a  strong  and  significant  negative effect of such payments on the increase of inequality as measured by the Gini index. When  analyzing each of the pillars separately, their individual effects on the decrease in inequality remain  significant,  particularly  for  pillar  2.  Further  disaggregating  each  of  the  pillars  by  their  payment  components gives no additional information on the particular performance of any of them in terms  of inequality. All the results described remain qualitatively similar when we use the Theil index as  our measure for inequality.  16   Table 11. CAP payments are associated with inequality reduction (a). Inequality measured by Gini Index (b). Inequality measured by Theil Index Notes: (1) Panel Fixed Effects including fixed effects; (2) Inequality is Notes: (1) Panel Fixed Effects including year fixed effects; (2) Inequality is measured by Gini index; (3) Data Source: Farm Structure Survey, Eurostat; measured by Theil index; (3) Data Source: Farm Structure Survey, Eurostat; EU-SILC, Eurostat; (4) Symbols: ** P<0.01, * P<0.05, +P<0.1; (5) Standard EU-SILC, Eurostat; (4) Symbols: ** P<0.01, * P<0.05, +P<0.1; (5) Standard errors are clustered at country level errors are clustered at country level Agriculture and poverty dynamics: Strategies at the household level    Besides  the  impact  of  the  CAP  funds  on  poverty  reduction,  other  analyses  were  performed  to  identify  additional  potential  factors  that  are  beneficial  to  poverty  alleviation.  In  particular,  the  following analysis focuses  on strategies at the  household level  that  have been associated  with a  reduction in poverty. We start by identifying certain agricultural activities that tend to be associated  to regions with higher poverty rates. The analysis performed shows that some agricultural activities  are  strongly  associated  to  regions  with  higher  poverty.  Table  12  below  records  the  spatial  correlation between certain agricultural activities and poverty. In particular, certain crops seem to  be the agricultural activity performed in poorer regions across the EU. Some of these crops include  specialist  horticulture,  specialist  vineyards,  combined  permanent  crops,  and  mixed  crops,  all  of  which show a significant association to poor regions. On the other hand, livestock activities tend to  develop in regions with lower poverty. Some of these include specialist dairying, specialist pigs, and  combined cattle dairying, rearing and fattening, all of which show a negative association to poor  regions.    17 Table 12. Relationship between agriculture area share by crop type and poverty Agriculture area with agriculture type: (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) LABELS Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Poverty rate Specialist horticulture - indoor 0.000548* Specialist horticulture - outdoor 0.000202** Specialist vineyards 3.07e-05** Various permanent crops combined 0.000188+ Specialist dairying -2.08e-05** Cattle-dairying, rearing and fattening combined -6.62e-05** Specialists pigs -1.07e-05* Various granivores combined 0.000168** Mixed cropping 6.65e-05** Mixed livestock, mainly grazing livestock 7.41e-05* Various crops and livestock combined 5.02e-05* Total utilised agricultural area 2.48e-06+ 2.44e-06* 2.22e-06* 1.80e-06* 5.79e-06** 3.62e-06** 2.60e-06+ 2.07e-06 1.77e-06** 2.93e-07 -7.74e-08 Population density(Inhabitants per hectare); 0.0987* 0.0959* 0.107* 0.106* 0.100* 0.108* 0.103* 0.0953* 0.0979* 0.0966* 0.0905* Gross Domestic Product(PPS per inhabitant) -0.000148* -0.000143+ -0.000154* -0.000153* -0.000142+ -0.000152+ -0.000153* -0.000147+ -0.000150* -0.000149* -0.000147* Poverty line -0.000356* -0.000431* -0.000402* -0.000435* -0.000385* -0.000351* -0.000408* -0.000327+ -0.000401* -0.000436+ -0.000389+ Dummy: agricultural data zero 1.674* 2.265+ 0.740 0.494 0.624 -0.167 0.563 1.777** 1.331+ 0.787 1.356 Dummy: data at NUTS 2 level -0.287 0.683 -0.774+ -0.428 0.309 -0.681+ -0.490 -0.657 -0.283 -0.711 -0.893+ Constant 22.56** 22.23** 23.49** 24.23** 23.38** 23.61** 23.96** 21.69** 23.33** 24.42** 23.64** Interaction termms with countries N N N N N N N N N N N Country fixed effects Y Y Y Y Y Y Y Y Y Y Y Observations 1,205 1,205 1,205 1,205 1,205 1,205 1,205 1,205 1,205 1,205 1,205 Adj R-squared 0.376 0.376 0.369 0.376 0.391 0.377 0.368 0.376 0.376 0.372 0.373 Note: (1) Data source: EU 2011 Poverty Map (DG-REGIO and World Bank), 2010 Farm Structure Survey (Eurostat), and EU National Statistic Institutes (Eurostat); (2) Symbols: ** p<0.01, * p<0.05, + p<0.1; (3) Missing observations in agricultural indicators are treated as zero; (4) Luxembourg and Cyprus Republic are not being analyzed here   In terms of poverty dynamics, we identified several strategies that have an impact on poverty over  time. Table 13 below summarizes two important results. First, that there is a positive association  between the poverty rate and the share of individuals who live in an agricultural household through  time. Thus, an increase in the share of individuals in agricultural households was associated with an  increase  in  poverty  through  time.  On  the  other  hand,  household  diversification  is  negatively  associated  with  the  poverty  rate  through  time.  Suggesting  that  as  diversification  of  the  income  sources at the household level increases, poverty decreases. Households that have more diverse  sources of income, in terms of agricultural and non‐agricultural income, are associated with a lower  poverty rate. This means that households who diversify seem to do better than those who rely only  on  agriculture.  Thus,  agricultural  households  could  benefit  from  diversifying  their  income  by  complementing it with non‐agricultural sources. However, it is important to note that this result  relies on well‐functioning labor markets, especially labor in the non‐agricultural sectors, that can  provide real alternative sources of income to members of agricultural households.      Table 13. Regions with higher household diversification have greater poverty reduction (1) (2) (3) (4) LABELS Poverty rate Poverty rate Poverty rate Poverty rate Share of individuals living in agriculture households 0.66* 0.24* 0.60* 0.42* Household diversification index -1.75* -1.07+ -0.29 GDP per inhabitant -10.57* -10.13* -8.46* Share of inhabitants with secondary education 0.16+ 0.15 0.04 Share of inhabitants with tertiary education -0.06 -0.06 -0.10 Unemployment rate 0.34+ R-Squared: Within 0.33 0.49 0.50 0.54 Between 0.21 0.14 0.16 0.24 Overall 0.25 0.14 0.16 0.24 Observations 959 959 959 959 18 Notes: (1) Panel Fixed Effects; All regressions include year fixed effects; (2) Anchored Relative Poverty line (60% of median at 2011); (3) Data Source: EU-SILC, Eurostat; (4) Symbols: ** P<0.01, * P<0.05, +P<0.1   Although diversifying the sources of income, between agricultural and non‐agricultural, may prove  useful to alleviate poverty in households over time, further results suggest that households who  receive agricultural income may be better off by specializing in a particular agricultural activity. Table  14  shows  that  an  increase  in  the  share  of  the  area  used  in  specialized  crop  production  seems  beneficial to poverty reduction, as regions with higher areas of specialization holdings show higher  poverty reduction. However, an increase in the share of land used for specialized livestock shows  no significant effect on poverty reduction. Hence, households who turn to agriculture as a source of  income do better when they specialize, particularly in crops. Thus, for poverty alleviation, it is useful  to  have  different  sources  of  income.  Nevertheless,  households  should  not  diversify  across  agricultural income sources.    Table 14. Regions with higher farm holdings specialization have higher poverty reduction (1) (2) (3) Poverty rate Poverty rate Poverty rate Agricultural area of specialized holdings share of total -0.02 -0.02 Agricultural area of holdings specialized in crop share of total -0.05+ Agricultural area of holdings specialized in livestock share of total -0.02 Ratio of agricultural area to total -0.03 -0.02 Agricultural employment share 0.22 0.22 0.22 GDP per inhabitant -0.01** -0.01** -0.01** Share of inhabitants with secondary education 0.28+ 0.28+ 0.30+ Share of inhabitants with tertiary education 0.16 0.16 0.17 Year fixed effects Y Y Y R-Squared: Within 0.42 0.42 0.43 Between 0.14 0.14 0.13 Overall 0.12 0.12 0.11 Observations 367 367 367 Notes: (1) Panel Fixed Effects; All regressions include year fixed effects; (2) Anchored Relative Poverty line (60% of median at  2011); (3) Data Source: Farm Structure Survey, Eurostat; EU‐SILC, Eurostat ; (4) Symbols: ** P<0.01, * P<0.05, +P<0.1    5. Poverty, agriculture, and the CAP: A comprehensive analysis    So far, we have documented how CAP funds can contribute towards inclusive growth by reducing  poverty and inequality. With this message in mind, the specific areas for improvement will depend  on a country’s level of structural transformation and on its current allocation of the CAP based on  high poverty areas. The goal of this section is to integrate the parts of the past analyses; on one side  we have the relationship between agriculture and poverty, which provides the state of a country  19 relative  to  a  successful  structural  transformation,  on  the  other  hand  we  have  the  relationship  between CAP and poverty, which identifies the regions which the CAP payments reach within the  country. In this section we integrate both parts, in order to identify areas for improvement. With  this in mind, we present the following figure, which plots the countries in terms of their association  between  poverty  rate  and  agricultural  indicators  (X‐axis),  and  their  association  between  poverty  rate and CAP payments (Y‐axis).     Figure 8. Country plot of the association between poverty rate, agriculture, and overall CAP payment  -6 -4 -2 0 2 4 6 80 Spain Romania Bulgaria 60 Average strength of association between CAP Portugal payments and Poverty (Standardized) Slovenia 40 Greece 20 East Germany Italy Malta Sweden 0 Belgium Finland UK Ireland Latvia Germany -20 Netherlands Austria France Denmark Poland -40 West Germany Slovakia Estonia -60 Hungary -80 -100 Strength of association between Agricultural share of area and Poverty (Standardized)     The results are easier to describe in terms of the plot quadrants. We start by describing the lower  left quadrant, which indicates a negative association both between poverty rate and agricultural  indicators, and poverty rate and CAP payments. This represents the case for successful structural  transformers, since it holds that agriculture is no longer associated to poverty and  thus the CAP  funds are not correlated to regions with high poverty. Some of the countries in this quadrant are  France, the Netherlands, Germany, Belgium, and Austria. In this case, we identify no specific areas  for improvement. Nevertheless, it is worth emphasizing that within this quadrant, we can find both  OMS and NMS. NMS in this quadrant suggests an efficient use of the CAP funding.    We  continue  describing  the  results  for  the  upper  right  quadrant,  which  indicates  a  positive  association in both indicators, implying that in these countries agriculture takes place in high poverty  regions and CAP funds reach the poorest regions. Despite the consistency in assigning CAP funds to  high poverty areas, and agriculture taking place in poverty areas as well, this is also an indicator that  this quadrant is located at the start of a structural transformation. For OMS countries located in this  quadrant, such as Spain, Portugal, Greece, and Italy, this incomplete structural transformation hints  at  the  existence  of  areas  for  improvement  to  achieve  an  efficient  use  of  CAP  funding.  These  countries may have been receiving the CAP for a longer time, and may thus have alternative ways  to use them more efficiently in order to achieve a successful transformation.     Finally,  the  lower  right  quadrant  underscores  the  most  potential  in  areas  for  improvement.  This  quadrant suggests that although agriculture takes place in poor regions within the country, the CAP  20 funds are reaching non‐poor regions instead. Countries in this group include Sweden and Latvia. In  this case, countries could improve poverty reduction by better targeting the CAP funds to include  poor regions where agriculture takes place.    6. Conclusion    In  this  paper  we  first  document  the  relationship  between  agriculture  and  poverty,  and  use  this  information to place the EU member states in their path towards achieving a complete structural  transformation. We then proceed to investigate the regions where the CAP funding has been going  within each country, by analyzing the relationship between the program’s payments and different  indicators of poverty. Additionally, we find that the CAP has contributed to poverty alleviation. In  this way, we conclude that the CAP is a powerful instrument towards achieving shared prosperity,  and  thus  also  complements  the  EU  goal  of  social  cohesion  by  reducing  disparities  in  the  regions  where agriculture prevails.    7. References    Crescenzi, R., & Giua, M. (2016). The EU Cohesion Policy in context: Does a bottom‐up approach  work in all regions? Environment and Planning A: Economy and Space 48(11), pp. 2340‐2357.     European Comission: Agriculture and Rural Development. (2017). The History of the Common  Agricultural Policy. Retrieved from: https://ec.europa.eu/agriculture/cap‐overview/history_en    European Parliament. (n.d.). Economic, Social, and Territorial Cohesion. Retrieved from:  http://www.europarl.europa.eu/atyourservice/en/displayFtu.html?ftuId=FTU_3.1.1.html    Mathur, S. K. (2005). Absolute convergence, its speed and economic growth for selected countries for 1961-2001. Journal of the Korean Economy, 6(2), 245-273.   World Bank (2018). Thinking CAP ‐ Supporting Agricultural Jobs and Incomes in the EU. EU Regular  Economic Report; no. 4. Washington, D.C. : World Bank Group.  http://documents.worldbank.org/curated/en/892301518703739733/EU‐Regular‐Economic‐ Report‐Thinking‐CAP‐Supporting‐Agricultural‐Jobs‐and‐Incomes‐in‐the‐EU    Sachs, J. D., & Warner, A. M. (1997). Fundamental sources of long-run growth. The American economic review, 87(2), 184-188.   United Nations Human Settlement Program. (2016). Structural Transformation in Developing  Countries: Cross Regional Analysis (Series 1). UN Habitat.         21 Appendix 1: Beta Convergence  We  measure  the  convergence  of  income  and  inequality  among  the  EU  regions  using  Beta  convergence. This method was developed by Sachs and Warner (Sachs and Warner 1997) following  the  Solow  Growth  Model.  This  tool  can  measure  the  trend  of  two  groups  of  regions  in  terms  of  economic growth, presenting the dynamic picture of their discrepancies. Old member states (OMS)  and new member states (NMS) were analyzed separately in order to show the income gap within  the EU.  Beta convergence is the result of regressing the annual growth rate of the analyzed variable on the  log of the variable from the previous year:   log ,   ℎ log , log ,   We use both the OLS regression and the fixed effects model (LSDV) to estimate the beta  convergence. The regressions are weighted using the average population in each region across the  analyzed period (2005 – 2014).     Table 1 shows the results on several income and inequality measurements that we selected. The  speed of convergence (β) is calculated as:     ln 1 100     Where T is the elapsed time, which equates 11 in this case.     The half‐life of convergence indicates how long it would take for the two groups of economies to  log 2 converge. Following the interpretations in Mathur (2005), it is roughly equal to  .    Appendix 2: Country specific effects    A key step in analyzing the countries’ stances in using the CAP fund is to calculate the country  specific strength of association between poverty and CAP as well as between poverty and  agriculture. We try to gain a comprehensive understanding on two issues: if the CAP funds are  going to the poor areas in a country and if a country’s agricultural activities take place mainly in  the poor areas.     In order to draw a picture of where each country stands, we use the 2011 NUTS 3 level data for  this analysis. The CAP funds distribution data are from the DG AGRI Clearance Audit Trail System  database (CATS), provided by the European Commission. The 2010 agricultural indicators from the  EU Farm Structure Survey are used to represent the regional agricultural situations. Since the  German agricultural data are only available at the NUTS 2 level, they are used as proxies for the  NUTS 3 level agricultural variables.     22 We use the regression model listed below to explain the regional poverty rates:         represents a series of variables including different categories of the CAP funds and normalized  agricultural indicators such as agricultural share of area. Population density is a proxy for the  urban rural typology.     For each variable used to explain the regional poverty, margins of responses for each country are  calculated to show the country‐specific strength of associations between poverty and the specific  variable. Tables 2, 3, 8, and 9 show the signs of the results, whereas the zeros indicate insignificant  marginal effects at 10% significance level. In Figure 8, the results were standardized by the  standard errors to become better comparable between variables.      Appendix 3: Analysis methods comparisons for CAP   Book Chapter 4 Our replication Our analysis 1 Our analysis 2 Data Outcome NUTS1/2 level NUTS3 level: NUTS3 level: NUTS1 and apportioned to GDP per head, Poverty rate, NUTS2 level: NUTS3 level: Population change, Share of the Poverty rate GDP per head, Poverty rate, Share of country poor Unemployment rate, the country poor Population change NUTS2 level: Unemployment rate CAP NUTS1/2 level NUTS3 level: NUTS3 level: NUTS1 and apportioned to CAP support per Total CAP NUTS2 level: NUTS3 level: hectare of utilizable support CAP support per CAP support per agricultural area, CAP capita hectare of utilizable support per agricultural area, CAP agricultural work unit support per agricultural work unit CAP data source FADN database; Clearance Audit Clearance Audit Clearance Audit appointed RD budgets Trail System (CATS) Trail System Trail System database (CATS) database (CATS) database CAP categories Pillar1: Market price Pillar1: Market price Pillar1: coupled Pillar1: coupled support and direct support and direct and decoupled and decoupled income payments income payments payments payments Method Pearson correlation Pearson correlation OLS regressions Fixed effects coefficients coefficients regressions Period 1999 2011 2011 2005-2013 Country coverage EU15 EU28 except EU28 except EU28 except Lithuania and Czech Lithuania and Germany Republic Czech Republic The table above shows comparisons of the different methods we adopted in order to study the correlation between poverty and the CAP support. The Book here refers to “The 23 CAP and the Regions: The territorial impact of the common agricultural policy” by Shucksmith, Thomson and Roberts. Following their work in the early 2000s, we attempt to re-examine social impacts of the CAP after its major reforms in 2003. More accurate CAP data become available for us in the CATS database, along with agricultural data at NUTS3 level from Eurostat’s Farm Structure Survey. In addition, more countries started receiving CAP funds. To capture this difference, we provide one set of results using all EU members and another set of results using only the EU15 in our replications of results of Shucksmith et al. Table 1. Pearson correlation coefficients between level of total Pillar 1 support accruing to NUTS3 regions (2008-2013) and socio-economic indicators (2011) Share of the Population Support per Unemploym country GDP per change hectare Support per Variables ent rate Poverty rate poor inhabitant (2008-2013) UAA AWU EU15 Support per hectare UAA 0.1163 0.0630* 0.1420** 0.1636** 0.0853** 1 0.6479** Number of observations 193 963 963 963 963 963 962 Support per AWU 0.0942 0.0846** 0.1540** 0.0651* 0.1954** 0.6479** 1 Number of observations 193 964 964 964 964 962 964 EU28 (except for Czech Republic and Lithuania) Support per hectare UAA 0.1061+ -0.0401 0.0353 0.0733+ -0.0125 1 0.5392** Number of observations 250 1165 1165 1165 1165 1165 1164 Support per AWU 0.0814 0.002 0.1195** 0.1098** 0.0628* 0.5392** 1 Number of observations 250 1166 1166 1166 1166 1164 1166 Note: (1) Symbol + indicates correlation statistically significant at 0.1 level; * indicates correlation statistically significant at 0.05 level; ** indicates correlation statistically significant at 0.01 level; (2) CAP support are the sum of payments for the period 2008-2013; (3) Unemployment rate data is at NUTS2 level; (4) EU15 are Belgium, Denmark, Greece, Finland, France, Germany, Ireland, Italy, Luxembourg, Malta, Netherlands, Portugal, Spain, Sweden and the UK Table 2. Pearson correlation coefficients between the level of market price support and direct income payments accruing to NUTS3 regions (2008 - 2013) and socio-economic indicators (2011) Share of the Population Support per Unemploy country GDP per change hectare Support per Variables ment rate Poverty rate poor inhabitant (2008-2013) UAA AWU EU15 Market Price Support Support per hectare UAA 0.0777 0.0252 0.0944** 0.1942** 0.0688* 1 0.6267** Number of observations 193 963 963 963 963 963 962 Support per AWU 0.0974 0.0436 0.1061** 0.1333** 0.1825** 0.6267** 1 Number of observations 193 964 964 964 964 962 964 Direct Income Payments Support per hectare UAA 0.1941** 0.1443** 0.1460** -0.0115 0.2126** 1 0.4994** Number of observations 193 965 965 965 965 965 964 Support per AWU 0.0567 0.0907** 0.0843** -0.1282** 0.2184** 0.4994** 1 Number of observations 193 966 966 966 966 964 966 EU28 (except for Czech Republic and Lithuania) Market Price Support Support per hectare UAA 0.0721 -0.0319 0.0469 0.1303** -0.0022 1 0.5748** Number of observations 250 1165 1165 1165 1165 1165 1164 Support per AWU 0.0904 0.017 0.0830** 0.1264** 0.3339** 0.5748** 1 24 Number of observations 250 1166 1166 1166 1166 1164 1166 Direct Income Payments Support per hectare UAA 0.1788** -0.0222 0.0326 0.0338 -0.0275 1 0.3095** Number of observations 250 1168 1168 1168 1168 1168 1167 Support per AWU 0.041 -0.0138 0.0261 -0.005 0.2042** 0.3095** 1 Number of observations 250 1169 1169 1169 1169 1167 1169 Note: (1) Symbol + indicates correlation statistically significant at 0.1 level; * indicates correlation statistically significant at 0.05 level; ** indicates correlation statistically significant at 0.01 level; (2) CAP support are the sum of payments for the period 2008-2013; (3) Unemployment rate data is at NUTS2 level; (4) EU15 are Belgium, Denmark, Greece, Finland, France, Germany, Ireland, Italy, Luxembourg, Malta, Netherlands, Portugal, Spain, Sweden and the UK; (5) Market price support is estimated to be all pillar 1 payments other than the direct aids payments Table 3. Pearson correlation coefficients between socio-economic indicators and the level of market price support accruing to NUTS3 regions, selected New Member States Share of Population Support No. Unemploy Poverty the GDP per change per Support observat ment rate rate country inhabitant (2008- hectare per AWU ions (NUTS 2) poor 2013) UAA Bulgaria Support per hectare UAA 28 -0.153 -0.1741 0.0779 0.9825** 0.8784** 1 0.9578** Support per AWU 28 -0.0784 -0.1304 0.052 0.9301** 0.7944** 0.9578** 1 Estonia Support per hectare UAA 5 N/A -0.579 0.1414 0.9959* 0.9374* 1 0.9890** Support per AWU 5 N/A -0.4609 0.0808 0.9779** 0.8830* 0.9890** 1 Hungary Support per hectare UAA 20 0.4716 -0.3448 0.1847 0.8639** 0.6857** 1 0.9998** Support per AWU 20 0.6152 -0.354 0.1756 0.8692** 0.6870** 0.9998** 1 Latvia Support per hectare UAA 5 N/A -0.4862 -0.3486 0.3241 0.4377 1 0.9959** Support per AWU 5 N/A -0.4236 -0.3479 0.2897 0.3666 0.9959** 1 Malta Support per hectare UAA 2 N/A -1 1.0000** 1.0000** 1.0000** 1 Support per AWU 2 N/A 1.0000** -1 -1 -1 -1 1 Poland Support per hectare UAA 47 -0.1704 -0.3893** -0.2216 0.5826** -0.1098 1 0.9945** Support per AWU 47 -0.1249 -0.3634* -0.2116 0.5306** -0.1008 0.9945** 1 Romania Support per hectare UAA 42 -0.1615 -0.4321** -0.1485 0.7534** -0.1508 1 0.9999** Support per AWU 42 -0.071 -0.4318** -0.1468 0.7556** -0.1504 0.9999** 1 Slov Republic Support per hectare UAA 8 -0.9443+ -0.6345 -0.6601 0.5403 0.2210 1 0.9978** Support per AWU 8 -0.9226+ -0.626 -0.6476 0.5315 0.2040 0.9978** 1 Slovenia Support per hectare UAA 12 1** 0.0704 0.2928 -0.0052 -0.1824 1 0.9929** Support per AWU 12 1** 0.0914 0.2839 -0.0535 -0.1975 0.9929** 1 Note: (1) Symbol + indicates correlation statistically significant at 0.1 level; * indicates correlation statistically significant at 0.05 level; ** indicates correlation statistically significant at 0.01 level; (2) CAP support are the sum of payments for the period 2008-2013; (3) Unemployment rate data is at NUTS2 level; (4) EU15 are Belgium, Denmark, Greece, Finland, France, Germany, Ireland, Italy, Luxembourg, Malta, Netherlands, Portugal, Spain, Sweden and the UK; (5) Market price support is estimated to be all pillar 1 payments other than the direct aids payments; (6) NMS not presented here are dropped due to the lack of observations 25 Table 4. Pearson correlation coefficients between total Pillar 2 support (2008-2013) and socio-economic indicators (2011) Share of the Population Support per Unemploy GDP per Support per Poverty rate country change hectare ment rate inhabitant AWU Variables poor (2008-2013) UAA EU15 Support per hectare UAA 0.1307+ 0.1132** 0.2213** 0.0905** 0.0831** 1 0.7018** Number of observations 193 963 963 963 963 963 962 Support per AWU 0.1076 0.1188** 0.1780** 0.0453 0.1097** 0.7018** 1 Number of observations 193 964 964 964 964 962 964 EU28 (except for Czech Republic and Lithuania) Support per hectare UAA 0.1121+ -0.0478 0.0183 0.0376 -0.0205 1 0.6478** Number of observations 250 1165 1165 1165 1165 1165 1164 Support per AWU 0.0829 -0.0186 0.1467** 0.0863** 0.0185 0.6478** 1 Number of observations 250 1166 1166 1166 1166 1164 1166 Note: (1) Symbol + indicates correlation statistically significant at 0.1 level; * indicates correlation statistically significant at 0.05 level; ** indicates correlation statistically significant at 0.01 level; (2) CAP support are the sum of payments for the period 2008-2013; (3) Unemployment rate data is at NUTS2 level; (4) EU15 are Belgium, Denmark, Greece, Finland, France, Germany, Ireland, Italy, Luxumbourg, Malta, Netherlands, Portugal, Spain, Sweden and the UK Table 5. Pearson correlation coefficients between level of LFA payment (2008-2013) and socio-economic indicators (2011) Share of the Population Support per GDP per Support per Unemploym Poverty rate country change hectare inhabitant AWU Variables ent rate poor (2008-2013) UAA EU15 Support per hectare UAA 0.0882 0.1219** 0.2913** -0.1447** -0.0404 1 0.4097** Number of observations 193 963 963 963 963 963 962 Support per AWU 0.0093 0.0771** 0.1162** -0.1108** -0.0221 0.4097** 1 Number of observations 193 964 964 964 964 962 964 EU28 (except for Czech Republic and Lithuania) Support per hectare UAA 0.0575 -0.0324 0.0568+ 0.0011 -0.0191 1 0.2781** Number of observations 250 1165 1165 1165 1165 1165 1164 Support per AWU -0.0071 0.0047 0.1760** -0.0517+ -0.0060 0.2781** 1 Number of observations 250 1166 1166 1166 1166 1164 1166 Note: (1) Symbol + indicates correlation statistically significant at 0.1 level; * indicates correlation statistically significant at 0.05 level; ** indicates correlation statistically significant at 0.01 level; (2) CAP support are the sum of payments for the period 2008-2013; (3) Unemployment rate data is at NUTS2 level; (4) EU15 are Belgium, Denmark, Greece, Finland, France, Germany, Ireland, Italy, Luxumbourg, Malta, Netherlands, Portugal, Spain, Sweden and the UK Table 6. Pearson correlation coefficients between level of agri-environmental subsidies (2008-2013) and socio-economic indicators (2011) Share of the Population Support per GDP per Support per Unemploy Poverty rate country change hectare inhabitant AWU Variables ment rate poor (2008-2013) UAA EU15 Support per hectare UAA 0.1572* 0.1570** 0.1796** -0.1043** 0.0201 1 0.4074** Number of observations 193 963 963 963 963 963 962 Support per AWU -0.0375 0.0489 0.0733* -0.0309 0.0792* 0.4074** 1 Number of observations 193 964 964 964 964 962 964 EU28 (except for Czech Republic and Lithuania) Support per hectare UAA 0.0933 -0.0385 0.0632* 0.0043 -0.0194 1 0.2800** Number of observations 250 1165 1165 1165 1165 1165 1164 Support per AWU -0.1093+ -0.0489+ 0.0903** 0.0212 0.0208 0.2800** 1 26 Number of observations 250 1166 1166 1166 1166 1164 1166 Note: (1) Symbol + indicates correlation statistically significant at 0.1 level; * indicates correlation statistically significant at 0.05 level; ** indicates correlation statistically significant at 0.01 level; (2) CAP support are the sum of payments for the period 2008-2013; (3) Unemployment rate data is at NUTS2 level; (4) EU15 are Belgium, Denmark, Greece, Finland, France, Germany, Ireland, Italy, Luxumbourg, Malta, Netherlands, Portugal, Spain, Sweden and the UK Table 7. Cross-tabulation of per hectare Pillar 1 support (2008-2013) and farm economic size (2010) Farm size classification (1 = smallest) Pillar 1 support per hectare classification (1 = lowest) 1 2 3 4 5 Total 1 0 7 103 48 75 233 0.00% 3.00% 44.21% 20.60% 32.19% 100.00% 2 36 31 70 47 49 233 15.45% 13.30% 30.04% 20.17% 21.03% 100.00% 3 50 54 31 55 43 233 21.46% 23.18% 13.30% 23.61% 18.45% 100.00% 4 56 58 19 44 56 233 24.03% 24.89% 8.15% 18.88% 24.03% 100.00% 5 72 87 20 39 15 233 30.90% 37.34% 8.58% 16.74% 6.44% 100.00% Total 214 237 243 233 238 1165 Table 8. Cross-tabulation of per hectare Pillar 2 support (2008-2013) and farm economic size (2010) Farm size classification (1 = smallest) Pillar 2 support per hectare classification (1 = lowest) 1 2 3 4 5 Total 1 0 3 108 59 63 233 0.00% 1.29% 46.35% 25.32% 27.04% 100.00% 2 3 13 73 61 83 233 1.29% 5.58% 31.33% 26.18% 35.62% 100.00% 3 23 61 24 62 63 233 9.87% 26.18% 10.30% 26.61% 27.04% 100.00% 4 74 78 21 38 22 233 31.76% 33.48% 9.01% 16.31% 9.44% 100.00% 5 114 82 17 13 7 233 48.93% 35.19% 7.30% 5.58% 3.00% 100.00% 27 Total 214 237 243 233 238 1165 Table 9. Agricultural factors influencing the level of CAP support Pillar 1 support per hectare Pillar 2 support per hectare β t β t Average farm size -10.45 -1.155 -5.766 -1.264 % holdings: rice -383.6 -1.164 -400.7 -1.145 % holdings: pasture -135.4* -2.216 -76.76 -1.356 % holdings: fruit -160.0 -1.429 -125.2 -1.105 % holdings: olives -92.13 -1.589 -79.11 -1.299 GDP per inhabitant 0.325 1.117 0.167 0.783 Number of employed persons 58.17 1.363 54.12 1.223 Population -0.00934 -1.085 -0.00673 -1.204 Population change -1.854 -0.908 -2.382 -1.028 Constant 552.0 0.0982 -890.6 -0.185     28