Transcript Diapositiva 1 - Regionale Economie Groningen

MASST – MAcroeconomic,
Sectoral, Social and
Territorial model
Topics and problems
Andrea Caragliu – Politecnico di Milano
Aims of the project
 The final goal of the project is forecasting future
socio-economic trends for European regions over a
period of 15 years from now.
 However, currently my commitment is to the
estimation stage.
2
Research steps
1.
2.
3.
Drawing up of a sound theoretical model and
definition of the appropriate econometric
counterpart;
Estimation of the model;
Forecast of main relationships and definition of
possible scenarios.
3
The MASST model - Logic scheme
4
Structure of the model
d r  d n  d r n
d r  f (Z n )  f (K r ,T r )
where:
Z = set of national demand variables
K = set of regional structural variables
T = set of regional territorial characteristics
5
The starting equation
 I use the following decomposition of regional growth rates:
yr  y n  s
where:
yr = variation in the region’s GDP
yn = variation in the nation’s GDP
s = shift
6
Estimated equations
I – National component
1 – GDP variation
Ynt   0  1Ct   2 I t   3Gt   4 X t   5 M t
where α = Parameters to be estimated
ΔC = Consumption growth rate
ΔI = Investment growth rate
ΔG = Public expenditure growth rate
ΔX = Exports growth rate
ΔM = Imports growth rate
7
Estimated equations
I – National component
2 – Consumption growth rate
Ct    cYt 1
3 – Public expenditure growth rate
Exogenous
8
Estimated equations
I – National component
4. Investment growth rate
I nt  Ynt 1  int 1  ULCnt 1  FDI nt 1
5. Export growth rate
ΔX nt = γ1 ΔULCt 1 + γ2 Ent 1
9
Estimated equations
II – Regional component
s = f (human and economic resources; structual and
sectoral characteristics; spatial spillover effects;
integration processes; territorial features)
10
New territorial data
Data
Definition
Source of raw data
Agglomerated regions
With a center of > 300.000 inhabitants and a population density >
300 inh./sq. Km. or a population density between 150 and 300 inh.
/sq. Km.
ESPON database
Urban regions
With a center between 150.000 and 300.000 inh. And a population
density of 150-300 inh./sq. Km. (or a smaller pop. density, 100-150
inh./sq. Km. with a bigger centre (> 300.000 inh.) or a population
density between 100 and 150 inh./sq. Km.)
ESPON database
Rural regions
With a population density < 100 inh./sq. Km. and centre > 125.000
inh. or a population density < 100 inh./sq. Km. with a centre <
125.000 inh.
ESPON database
Megas regions
Regions with the location of at least one of the 76 FUAs with the
highest average score in a combined indicator of transport,
population, manufacturing, knowledge, decision-making in the
private sector
ESPON database
Pentagon regions
Regions located within the Pentagon formed by the five European
cities of Milan, Munich, Amsterdam, London, Paris
ESPON database
11
New socio-economic data
Data
Definition
Source of raw data
Regional energy consumption by
population
Total energy consumption on population at NUTS 2 in the
year 2002
ESPON database
Net immigration flows (people
between 17 and 27 years)
Average immigration flows of people between 17 and 27
years in the period 1/1/95 - 1/1/00 at NUTS 2 level
ESPON database
Net immigration flows (people
between 32 and 42 years)
Average immigration flows of people between 32 and 42
years in the period 1/1/95 - 1/1/00 at NUTS 2 level
ESPON database
Net immigration flows (people
between 52 and 67 years)
Average immigration flows of people between 52 and 67
years in the period 1/1/95 - 1/1/00 at NUTS 2 level
ESPON database
Regional birth rate
Share of births on population at NUTS 2 level
ESPON database
Regional mortality rate
Share of deaths on population at NUTS 2 level
ESPON database
12
Spatial effects indicators
Indicators
Definition
Source of raw data
Spatial spillovers
Sum of the relative annual growth rates of all regions other than region i
divided by the distance between each other region and region i.
Eurostat
Economic potential
Sum of the annual absolute difference between income growth rates of
region j and region i divided by the distance between region i and all
other regions j.
Eurostat
Integration potential
Change in the sum of the annual absolute difference between income
growth rates of regions j and region i divided by the distance between
region i and all other regions j, when in the second term distance is
squared for those regions at the border between Eastern and Western
Countries.
Eurostat
13
Traditional economic variables
National variables
Defintions
Sources of the raw data
GDP growth rate
Annual % growth rate of real GDP at NUTS 0 in
the period 1995-2002
Eurostat
Annual change in interest rate
Absolute change in short-term interest rates (3
months) at NUTS 0 in the period 1995-2002
Eurostat
Annual change in unit labour cost
Absolute change in unit labour cost (calculated as
unit salary * number of employees / GDP) at
NUTS 0 in the period 1995-2002
Eurostat
Share of FDI on total internal investments
% Flow of FDI / Gross Fixed Capital Formation at
NUTS 0 in the period 1995-2002
Eurostat
Nominal exchange rate
Nominal effective exchange rate at NUTS 0 in the
period 1995-2002
Eurostat
Inflation rate
Inflation rate at NUTS 0 in the period 1995-2002
Eurostat
Consumption growth
% annual real consumption growth rate at NUTS 0
in the period 1995-2002
Eurostat
Investment growth
% annual real gross fixed capital formation growth
rate at NUTS 0 in the period 1995-2002
Eurostat
Import growth
% annual real import growth at NUTS 0 in the
period 1995-2002
Eurostat
Eastern Countries
All former Eastern Economies
New EU Countries
The 10 new Member Countries who joined the EU
on the 1/5/04
14
Population growth rate
Pt 1  0  1 fr  2mr  3im
where
fr = fertility rate - exogenous
mr = mortality rate - exogenous
im = interregional migration  im  0  1u  2 ( we  wr )
where
u = unemployment
we = European average wage
wr = regional average wage
15
Database and indicators
 The database is built for 27 Countries (all EU25 countries
plus Bulgaria and Romania) and 259 regions (NUTS2). The
national database is in panel form (1995-2002).
 The database’s originality is due to:
- The use of territorial and socio-economic data at NUTS2
level (so far inexistent), coming from other ESPON projects;
- The use of other spillover indicators created for 259 regions;
- Building up a database which is consistent with Eurostat and
ESPON sources for which missing values were filled and
consistency was checked.
16
Results of estimation of shift
parameters
17
Open questions 1 - Econometrics
1. As I am estimating spatial spillover effects, most
of the spatial autocorrelation should be already
wiped out. Which kind of spcification test, in the
shape of the Moran’s I, might I use in this case?
18
Open questions 1 - Econometrics
2.
The spillover equation can be written as
r
y j
D
i 1
, i  j
i, j
Therefore, I am already using income in the equation. Am I
running into endogeneity of the regressors problem?
19
Open questions 1 - Econometrics
3. Regional shift effects do not automatically sum up
to 0 (as we would wish for); instead, given the fact
that the describing equation is filled with positive
explanatory variables, they tend to be distorted
towards positive values. Summing up to 0 is
imposed in the estimation process; is there any
alternative solution?
20
Open questions 2 - Economics
1. Calculated shift s, plotted for each year and each
region, is characterized by high variane. That’s why
its average over the period 1999-2002 is chosen.
This choice should be econometrically correct, bu
how do I motivate it from the theoretical point of
view?
21
Open questions 2 - Economics
2. Again from the theoretic point of view, why is σ2s
so high?
22
Open questions 2 - Economics
3. In the national equations subgroup, consumption
growth rate was described by the following
expression:
Ct    cYt 1
It is in reduced form, which is a technique used in all
the equations. Given its econometrically accetable
use, how do I justify it from the economic
perspective?
23