1. Spatial and non spatial approaches to agricultural convergence in Europe
Luciano Gutierrez*, Maria Sassi**
*University of Sassari **University of Pavia

2. The role of spatial effects
Political and financial perspective
Empirical perspective
1. Introduction
- Real convergence: a key objective of the EU
- Little attention
- Interest to agriculture
- Rather small number of studies that deal with theoretical and empirical advancement
Accelleration of growth and income
CAP and RD for territorial disparities reduction
The role of spatial effects

3. 2. Cross-sectional regressions
1. Introduction
1. Barro-style methodology
Spatial effects
2. Outline
3. Panel data regressions
80 EU regions
NUTS2
( / )

4. 1. Barro-style methodology
Per capita income at the initial year
1. Introduction
2. Outline
3. Cross-sectional models
Annual average growth rate of per capita income
parameter of convergence
If b is negative and statistically significant, the neoclassical hypothesis of convergence is verified: a process only driven by the rate of technological progress

5. Technological diffusion Neoclassical perspective
1. Introduction
Neoclassical perspective
Economic geography
2. Outline
3. Cross-sectional models
knowledge
entirely disembodied and understood as a pure public good
a regional public good with limited spatial range
by Anselin (1988) a growing literature has shown the relevance of the problem and the errors and misspecifications that can occur if the issue is ignored in cross-sectional data analysis involving geographic units.
differentials of income and growth rate across regions cannot be explained in terms of different stocks of knowledge
regions might show different path of growth even in opposite direction
Empiric literature

6. Spatial effects 1. Spatial autocorrelation 2. Spatial heterogeneity
1. Introduction
2. Outline
1. Spatial autocorrelation
2. Spatial heterogeneity
3. Cross-sectional models
Coincidence of attribute similarity and location similarity
Assumption of independent residuals
Unobservable variables and steady state
The value of variables sampled at nearby location are not independent from each others
Geographic spill-over effects

7. 1. Barro-style methodology Endogenous spatial lag variable
1. Introduction
2. Outline
3. Cross-sectional models
2. Spatial lag model
Endogenous spatial lag variable
3.1 Global spatial cross-sectional models
3. Spatial error model
Omitted variables
Ro is the spatial autoregressive parameter that indicates the extent of interaction between observations according to a spatial pattern exogenously introduced by means of the standardized weight matrix W.

8. 1. Spatial autocorrelation 2. Spatial heterogeneity
Spatial effects
1. Introduction
2. Outline
1. Spatial autocorrelation
2. Spatial heterogeneity
3. Cross-sectional models
Structural instability or group-wise heteroskedasticity
Convergence clubs
Possibility of multiple, locally stable steady state equilibria

9. 1. Barro-style methodology
1. Introduction
2. Outline
3. Cross-sectional models
4. GWR models
3.1 Global spatial cross-sectional models
3.2 Local spatial cross-sectional models
each data point is a regression point that is weighted by the distance from the regression point itself

10. 1. Barro-style methodology c. Spatial autocorrelation
2. SLMs SEMs
1. Introduction
2. Outline
3. Cross-sectional models
5. Panel data models
3.1 Global spatial cross-sectional models
3.2 Local spatial cross-sectional models
4. Panel data models
a. Time dependence
c. Spatial autocorrelation
b. Space dependence

11. 5. Panel data models Serial dependence of the dependent variable
1. Introduction
2. Outline
3. Cross-sectional models
3.1 Global spatial cross-sectional models
Serial dependence of the dependent variable
Space-time autoregressive and space-time dependence
3.2 Local spatial cross-sectional models
4. Panel data models
Intensity of the contemporaneous spatial effect

12. 5. Panel data models
1. Introduction
2. Outline
3. Cross-sectional models
3.1 Global spatial cross-sectional models 1.
3.2 Local spatial cross-sectional models
Dependence results from the neighborhood locations in the previous time period
4. Panel data models

13. 5. Panel data models
1. Introduction
2. Outline
3. Cross-sectional models
3.1 Global spatial cross-sectional models 1.
3.2 Local spatial cross-sectional models 2.
4. Panel data models
Dependence results from location and its neighborhood in the previous time period

14. Time and spatial lag are included
5. Panel data models
1. Introduction
2. Outline
3. Cross-sectional models
3.1 Global spatial cross-sectional models 1.
3.2 Local spatial cross-sectional models 2.
4. Panel data models 3.
Time and spatial lag are included

15. 5. Panel data models 1. 2. 3. 4. 1. Introduction 2. Outline
3. Cross-sectional models
3.1 Global spatial cross-sectional models 1.
3.2 Local spatial cross-sectional models 2.
4. Panel data models 3. 4.

16. 5. Panel data models 1. 2. 3. 4. 5. 1. Introduction 2. Outline
3. Cross-sectional models
3.1 Global spatial cross-sectional models 1.
3.2 Local spatial cross-sectional models 2.
4. Panel data models 3.
Concerning the spatial arrangement W is based on euclidian distances among regions row-normalised and with W that remains constant over time. 4. 5.

17. 5. Panel data models GMM estimator
1. Introduction
2. Outline
3. Cross-sectional models
3.1 Global spatial cross-sectional models
All the special cases of the general specification can be estimated with only few modifications to moment restrictions
3.2 Local spatial cross-sectional models
GMM estimator
4. Panel data models
With spatial lags it shows good propertis and can be estimated easly

18. Barro-style methodology
Results
Barro-style methodology
1. Introduction
2. Outline
3. Cross-sectional models
3.1 Global spatial cross-sectional models
3.2 Local spatial cross-sectional models
4. Panel data models
4. Results

19. Results 1994-2007 1. Introduction 2. Outline 3. Cross-sectional models
3.1 Global spatial cross-sectional models
3.2 Local spatial cross-sectional models
4. Panel data models
4. Results

20. Results: GWR 1. Introduction 2. Outline 3. Cross-sectional models
3.1 Global spatial cross-sectional models
3.2 Local spatial cross-sectional models
4. Panel data models
4. Results

21. Results: GWR and local parameters
1. Introduction
2. Outline
3. Cross-sectional models
3.1 Global spatial cross-sectional models
3.2 Local spatial cross-sectional models
4. Panel data models
4. Results

22. GWR and local parameters of convergence (1994-2007)
1. Introduction
2. Outline
3. Cross-sectional models
3.1 Global spatial cross-sectional models
3.2 Local spatial cross-sectional models
4. Panel data models
4. Results

23. Results: dynamic spatial panel model (1980-2007)
1. Introduction
2. Outline
3. Cross-sectional models
3.1 Global spatial cross-sectional models
3.2 Local spatial cross-sectional models
4. Panel data models
4. Results

24. Results: dynamic spatial panel model (1994-2007)
1. Introduction
2. Outline
3. Cross-sectional models
3.1 Global spatial cross-sectional models
3.2 Local spatial cross-sectional models
4. Panel data models
4. Results

25. Conclusions
Little formal guidance available for cross-country and panel data spatial models (Florax & de Graaff)
Specification of the weight matrix
1. Introduction
2. Outline
3. Cross-sectional models
Global spatial cross-sectional models
Spatial panel data models
3.1 Global spatial cross-sectional models
Exogenous constructed W matrix
3.2 Local spatial cross-sectional models
Binary scheme designed according to the Queesn’s contiguity
Euclidian distances – row normalised
4. Panel data models
Space has a role in convergence but it is strongly dependent on the specification of the W matrix
4. Results
Fixed vs. adaptive bandwidth
4.1 Cross-sectional models
Enogenous constructed W
GWR
4.2 Panel models
n. regions into the kernel
5. Conclusions
Type of spatial weight

26. Conclusions Convergence clubs Panel data environment?
1. Introduction
Convergence clubs
Panel data environment?
2. Outline
3. Cross-sectional models
3.1 Global spatial cross-sectional models
Different explanations by theoretical literature
3.2 Local spatial cross-sectional models
Neoclassical perspective
Endogenous growth th.
4. Panel data models
Space has a role in convergence but it is strongly dependent on the specification of the W matrix
4. Results
4.1 Cross-sectional models
Saving rate out of wage larger than saving rate out of capital
Different initial values of human capital and knowledge
4.2 Panel models
5. Conclusions

27. Conclusions Regions with equilibrium values below the average
1. Introduction
Regions with equilibrium values below the average
2. Outline
Spatial autocorrelation and heterogeneity
3. Cross-sectional models
3.1 Global spatial cross-sectional models
Policy interventions
3.2 Local spatial cross-sectional models
NUTS2 Administrative units
Different agricultural and socio economic regions
4. Panel data models
Space has a role in convergence but it is strongly dependent on the specification of the W matrix
4. Results
4.1 Cross-sectional models
4.2 Panel models
5. Conclusions