Content uploaded by Tommy Lundgren
Author content
All content in this area was uploaded by Tommy Lundgren
Content may be subject to copyright.
1
The Determinants of Economic Growth in the Swedish
Mountain Region – the Role of the Forest and Tourism Sector, and
Protected Land
Tommy Lundgren
Department of Forest Economics
SLU
S-901 83 Umeå
Sweden
Email: tommy.lundgren@sekon.slu.se
Phone: +4670 – 517 4396
Fax: +4690 – 786 6073
Abstract
This paper investigates the determinants of economic growth and growth patterns in the Swedish
mountain area municipalities. Focus is on the effects of net migration, the forest sector, tourism, and
protected land. We use a standard panel data setting with data containing 15 municipalities spanning
the years 1985 to 2001. System equation GMM estimation results show that: 1) absolute convergence
may be applicable (“poorer” regions grow faster than “rich” regions), but there is no evidence for
conditional convergence among the mountain municipalities (additional assumption: different steady-
states across economies); 2) forest industry employment have a positive impact on local economic
growth and vice versa; 3) tourism does not have a significant effect, either positive or negative, on local
growth and vice versa; 4) the development of national GDP is important for local economic growth
(+), net migration (-), the forest sector (-), and tourism (+); 5) amount of protected land in relation to
toal land area does not seem to be important for the local mountain economy, except for tourism, where
it has a statistically weak negative impact.
Keywords: convergence, GMM, growth theory, empirical growth modeling
2
1 Introduction
The main objective of this paper is to study determinants of economic growth in the
Swedish mountain area municipalities during the period 1985-2001. The main
objective can be divided into three distinct sub-objectives: (i) convergence patterns in
economic development will be assessed (absolute and conditional); (ii) forest sector
activity, tourism sector activity and their impact on economic growth will be
analyzed; (iii) the effect of protected land on regional economic growth, the forest
sector, and tourism, is evaluated.
The Swedish government’s decision to increase the amount of protected land, to some
extent that will include productive forest land, has created a lively debate of how it
will affect communities which are more or less dependent on income from forestry
related activities. A fundamental question is whether such policy will hamper
economic growth in these communities. Opponents of environmental protection claim
that protected lands limit growth opportunities in adjacent communities by locking up
potentially valuable natural resources and restricting extractive industries and other
business activities. Proponents claim that extractive industries, such as logging and
mining, are no longer the backbone of rural economies – instead, the presence of
protected lands encourages growth by attracting tourists and potentially new residents
through positive net migration. Studies in the US, for example Lorah and Southwick
(2003) and Rasker and Alexander (2003), show that protected land have a positive
effect on economic development in nearby communities. Jonsson (2004), using a
standard growth model à la Barro and Sala-i-Martin (2004) with gross regional
production as dependent variable, find that protected land in a part of the Swedish
mountain area (8 municipalities) does not have a significant effect on growth during
the period 1985-2001. In a study of economic growth and the amount of wilderness
3
land in Norway, Skonhoft and Solem (2001) find a negative relationship; that is, high
economic growth is associated with less wilderness land.
The effects of policy aimed at protecting the environment on forestry and local
economies have been assessed by Berck et al (2000). Using a multi-county time series
approach (VAR) to model the economy in timber-dependent counties in California,
they find that there is no evidence that such policy has hampered local economic
growth. Instead, it is nation-wide and federal economic growth that matters for local
growth and poverty. Jonsson (2004) also try to measure the effect of forest sector
business activity (in terms of timber harvesting) on local economic growth in a part of
the Swedish mountain region. His results show that, during 1985-2001, there is no
evidence that forestry activity (harvesting) have had a positive effect on local
economic growth (measured as annual percentage growth in gross regional product).
The empirical literature on economic growth is vast and includes studies with data
sets that are country specific, country panels, complete region panels for countries,
and incomplete region panels within countries. The latter, which this study belongs to,
is not very common. Especially when it comes to sparsely populated sub-urban areas
like the municipalities in the Swedish mountain region. Traditionally, empirical
analysis of economic growth is performed with the help of a single growth equation,
or in some cases a growth equation and a net migration equation (see, for example,
Barro and Sala-i-martin, 2004, or Lundberg, 2001). Furthermore, Swedish studies
focus usually on the issue of convergence and fiscal or other public/political matters
(see, for example, Eliasson and Westerlund, 2003, Aronsson et al, 2001, Lundberg,
2001, or Persson, 1997). In this paper, however, we add two equations to this setting;
one for the demand for forest sector labor, and one for the demand for tourism sector
4
labor. As far as we know, this multi-equation approach is novel in empirical growth
modeling.1
The paper is organized as follows. Section 2 briefly describes the theory behind the
empirical specifications. Section 3 presents some stylized facts from the region;
growth rates, net migration, employment in tourism and the forest sector, and amount
and development of protected land. Section 4 is devoted to empirical specifications,
analysis, and results. Section 5 concludes.
2 Theoretical background
Barro and Sala-i-Martin (2004) and Aghion and Howitt (1998) outline and summarize
the theoretical - and to some extent the empirical - work on economic growth. The
work span from the neoclassical models with exogenously determined growth, to
models with growth generated endogenously developed in the last two decades. Many
empirical papers are aimed at testing the neoclassical theory, which predicts
convergence in growth across regions, i.e., small economies grows faster than large
economies.
In this section we briefly outline the basic structure of neo-classical growth theory.
We believe this is pertinent to the subsequent analysis, since the issue of absolute and
conditional convergence will be dealt with in the empirical application. Growth
theory proposes that to sustain a positive growth rate of output per capita, in the long-
run, continual augmentation of the stock of technological knowledge (technological
progress) is necessary (Solow, 1956, Swan, 1956). Assume that aggregate production
depends on capital and labor according to a CRTS Cobb-Douglas production
function:
1 To our knowledge, only a few studies, such as Lundberg (2001) and Aronsson et al (2001)
estimate growth and net migration equations simultanously.
5
()
,Κ== −1 aa
LLKFY ,
with diminishing returns,
.0,0,0,0 2
2
2
2
<><> dL
Fd
dL
dF
d
K
Fd
dK
dF
Everybody in the economy supplies one unit of labor per unit of time and there is
perpetual full employment. This means L equals the population. The population grows
at rate n. In this setting, output per person, Y/L=y, depend on capital stock per person,
K/L=k. Then per capita production can be written,
.== a
kkfy )(
Capital evolves over time according to,
()
,−= dKLKsF
dt
dK ,
()
,, ILKsF =
where s is the rate of saving in the economy, and I is gross investment in capital. In
per capita terms, the capital equation of motion is,
.)+(−=)+−= knskknksf
dt
dk a
δδ
()(
6
Note that the differential equation governing the capital-labor ratio is almost the same
as the differential equation for capital. F(K, L) is replaced by f(k), and the depreciation
rate is augmented with the population growth, n. Faster growth of the population will
tend to reduce the amount of capital per capita in the same way as depreciation would.
Under CRTS the absolute size of the economy, K, is irrelevant. Instead, what matters
is the relative factor production, k.
Figure 1. The Solow-Swan neoclassical growth model
Diminishing returns impose an upper limit to capital per person. A point will be
reached where all saving is needed to compensate for depreciation and population
growth; the steady state. The steady-state capital stock, k*, is given by,
,)+(= knsk a
δ
dk/dt
k(0) k* Capital per person
Saving per person
Depreciation and dilution per person
Depr + dilution per
person
Saving per person
= sf(k)
kn )(
δ
+
=
7
.
)+(
=1−
1
a
s
n
k
δ
*
The corresponding steady state level for output per capita is given by
(
)
** kfy =. In
steady state, output and capital will grow at the rate of population growth. Growth as
measured by the rate of increase in output per person will stagnate and cease in the
long-run. This model feature does not translate well into reality, where growth rates
seem to be positive even in the long-run.
One way of dealing with diminishing returns is to assume exogenous technological
change. Assume a productivity parameter A in the production function that reflect
current state of technology, and that A grows at the rate g. The exogenously given rate
g reflect progress in science. Then we have,
()
.= −1 a
aKALY
Writing the aggregate output this way makes technological progess the same as
increasing the ”effective” labor force, AL, which grows at rate g + n. The only
difference from previous model is that we substitute labor L for ”effective” labor AL,
and we raise the ”effective” population growth from n to g + n. By the same
reasoning as before, we see that capital and output per capita will approach a steady
state. However, in this steady state, output and capital grows at the same rate as the
effective population, AL. This means that per capita output and capital, in steady state,
will grow at the exogenous rate of technological progress, g.
Consider two countries or regions with the same technologies, and the same values of
the parameters that determine the steady state. Then, the country or region that begins
8
with a low level of capital or output per capita must have a higher growth rate. Since
a
ky =, we can write
,
//
k
dtdk
a
y
dtdy =
and they share the same value of a. Therefore the lagging country (with the lower k)
must have a faster growth rate of output per capita. What this means is that the longer
k is from k*, the higher the growth rate of capital per capita. This is referred to in the
literature as conditional convergence. Convergence is often tested in the literature
using the following equation;
itit
it
Tit ybx
y
y
T
ε
++)(−=
+
it
γXloglog
1
The equation says; growth rates can vary from country to country either because of
differences in variables determining their steady states, the term it
γX, or because of
differences in initial positions, captured by the term )
(
−
it
yblog . An estimated value
of b > 0 is taken as evidence for conditional convergence. An annual period-to-period
model would be formulated as;
Here we would assume that economy i reach steady-state every period (in some
sense). The above equation is our outset when conducting the empirical analysis. But
itit
it
it ybx
y
y
ε
++)(−=
−−
−
1it
γX
1
1
loglog
9
first we present some stylized facts about economic growth, forest sector activity,
tourism, and protected land.
3 Some stylized facts from the region
This study focus on the 15 Swedish mountain municipalities during the period 1985 to
2001. From north to south, these are; Kiruna, Gällivare, Jokkmokk, Arjeplog, Sorsele,
Storuman, Vilhelmina, Dorotea, Strömsund, Krokom, Åre, Berg, Härjedalen,
Älvdalen, and Malung. From these municipalities we collect a set of cross-section-
time-series variables each consisting of 255 observations in total (a panel data set).
The first problem one encounter when looking at economic growth in a region is what
measure to use. Many possible variables could measure welfare; consumption, total
production value, different types of income measures, etc. In figure 2 the total per
capita income from employment is depicted for the years 1985 and 2001. It is obvious
that there is variation across municipalities, with Kiruna and Gällivare showing the
largest income levels, and that all show a positive growth in income between 1985
and 2001.
10
Income per capita 85/01
0
10
20
30
40
50
60
70
80
90
100
Kiruna
Gällivare
Jokkmokk
Arjeplog
Sorsele
Storuman
Vilhelmina
Dorotea
Strömsund
Krokom
Åre
Berg
Härjedalen
Älvdalen
Malung
SEK*1000
Y_85
Y_01
Figure 2. Income from salary of municipality residents (per capita) 1985 and 2001
Figure 3 show per capita gross regional product (GRP) for all regions. Here we can
see that there is greater variation between regions than in the case of income, and
Jokkmokk show relatively larger production values compared to the other regions.
This is due to the substantial amount of hydropower production in Jokkmokk. All
municipalities, except Åre, has had a positive growth rate of GRP between 1985 and
2001. Why the difference between income and gross regional product? First, gross
regional product is measured as value added; that is, the difference from the
employment income measure is income from capital, which in most cases would
make per capita GRP different from per capita inome. Second, there is also a
difference in how the variables are measured; employment income being measured as
income of municipality residents, and GRP being measured as value added for the
firms located in the region. This means that if a firm suddenly move their HQ from
Kiruna to, say, Luleå, then large production values will emigrate from the region. In
11
this case, it will show up in the data as production value losses, even though the main
production activitiy is still located in the region. Therefore, we have chosen not to use
this measurement as an indicator of regional welfare. Instead we focus on per capita
employment income.
GRP per capita 85/01
0
100
200
300
400
500
600
Kiruna
Gällivare
Jokkmokk
Arjeplog
Sorsele
Storuman
Vilhelmina
Dorotea
Strömsund
Krokom
Åre
Berg
Härjedalen
Älvdalen
Malung
SEK*1000
GRP_85
GRP_01
Figure 3. GRP per capita in 1985 and 2001.
In figure 3 we depict the regional growth rates between 1985 and 2001 for per capita
employment income and per capita GRP. For the whole mountain region; average
growth rate of GRP is 1.31%, and the average growth rate of income is 1.60%.
Compared to the growth rate for national GDP, 1.59%, we can not find any evidence
that the mountain region has lagged behind the rest of Sweden during 1985-2001. At
least not if economic growth is measured as growth in per capita income from
employment. In this figure it is obvious that the two different measures tells us
“different stories”, and since per capita employment is the more reliable measure, in
terms of adequately measuring regional economic activity, we will only focus on this
variable as an indicator of economic welfare from now on.
12
Growth rates GRP and income 1985-2001
-0,5
0
0,5
1
1,5
2
2,5
3
Kiruna
Gällivare
Jokkmokk
Arjeplog
Sorsele
Storuman
Vilhelmina
Dorotea
Strömsund
Krokom
Åre
Berg
Härjedalen
Älvdalen
Malung
Percent
Figure 4. Annual growth rates for GRP and income from employment 1985 to 2001.
About 2% of Sweden’s population live in the mountain region and the population
density is 1.2 persons per km2. Sweden’s total population is 9 million, which
translates into about 20 persons per km2. The mountain region is suffering from
depopulation and an aging population. Net migration flows are described in table 1,
and employed in the forest sector and tourism sector in table 2. The figures show that
there has been considerable negative net migration across all mountain municipalities
(except Berg and Krokom where quite a few years are characterized by positive net
migration). In the time interval 1985-2001, employed in the forest sector has
decreased by 50%, while employed in tourism is virtually unchanged.2 Some of the
decrease in the forest sector is due to technical development in forestry, but a
significant part is due to lower harvesting and other activities related to the forest
sector.
2 See Appendix 1 for details on sub-sectors included in the tourism sector (SNI-codes).
13
Table 1. Net migration in all mountain municipalities, 1985-2001.
1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
Malung -46 11 -5 48 95 -31 -5 -33 -10 102 -122 -24 -95 -115 -1 -82 -78
Älvdalen 6 16 11 38 -4 101 19 -27 -37 20 -7 -28 -19 -94 -73 -37 -2
Krokom -58 -97 33 113 77 258 209 122 76 60 -138 1 -81 -208 2 -36 -64
Strömsund -55 -133 -101 44 50 -24 -24 -89 -38 -45 -110 -92 -133 -104 -172 -159 -165
Åre 64 -58 -1 204 77 81 -13 -33 58 113 11 -18 -138 -100 -40 19 -119
Berg -15 -49 49 20 80 90 83 -3 -7 3 21 -9 16 -23 59 -42 -47
Härjedalen -118 -66 2 130 -3 -61 -26 -2 -6 -38 -22 -52 -77 30 -37 -109 -3
Storuman -75 -92 -115 -94 -44 -102 -16 15 -115 -97 -9 -48 -36 -96 -71 -77 -92
Sorsele -69 -21 -26 -54 20 -1 -15 -5 5 -34 1 -14 5 -24 -9 -42 -37
Dorotea -16 10 8 -26 6 9 5 24 40 -122 10 -66 -13 -34 -38 19 -49
Vilhelmina -52 -55 -92 0 39 -23 -59 -11 62 42 -104 -16 -22 -106 -46 -57 -138
Arjeplog -3 -100 -21 2 13 -48 -3 0 16 -39 34 -64 -30 -38 -23 -51 -39
Jokkmokk -12 -48 -10 24 -9 -99 -75 -51 -2 -19 -30 20 -63 -73 -103 -70 -42
Gällivare -234 -301 -411 -313 -249 -215 -149 -89 29 -250 -276 -288 -290 -358 -419 -437 -289
Kiruna -457 -249 -292 -370 -156 -417 -173 -144 -56 -270 -436 -274 -348 -172 -361 -511 -494
14
Table 2. Employed in the forest and tourism sector, 1985/2001.
Figure 5 describe the amount of protected land in each municipality (in
relation to total land area) for 1985 and 2001. There has been a considerable
accumulation of protected land during the period; some municipalities
Forest industry and tourism sector employed 1985/2001
Forest industry Tourism sector
All mountain municipalities 11417/5328 5010/5161
Norrbotten 2496/835 1736/1397
Västerbotten 2242/998 498/645
Jämtland 4877/2505 1908/2062
Dalarna 1802/990 868/1057
Municipalities
Kiruna 679/335 859/704
Gällivare 949/263 596/388
Jokkmokk 503/174 166/189
Arjeplog 365/63 115/116
Sorsele 479/197 51/50
Storuman 728/273 191/299
Vilhelmina 638/320 203/239
Dorotea 397/208 53/57
Strömsund 1575/866 207/188
Krokom 831/475 163/273
Åre 527/225 774/770
Berg 663/398 164/188
Häjedalen 1281/511 600/643
Älvdalen 952/480 386/444
Malung 850/510 482/613
15
starting at 0% and ending at 20-20%. Protected land in Storuman and
Jokkmokk constitutes about 50% of total land. The question is: can we see
any effects of the accumulation of protected land on economic growth in the
mountain region? As the empirical analysis will reveal, we can not.
Protected land in relation to total land area 85/91
0
0,1
0,2
0,3
0,4
0,5
0,6
Fjällkommunerna
Norrbotten
Västerbotten
Jämtland
Dalarna
Kiruna
Gällivare
Jokkmokk
Arjeplog
Sorsele
Storuman
Vilhelmina
Dorotea
Strömsund
Krokom
Åre
Berg
Häjedalen
Älvdalen
Malung
Figure 5. Protected land in relation to total land area, 1985 and 2001.
Figure 6 shows how initial amount of protected land (1985) and subsequent increases
in protected land are related. It seems like municipalities with low protected land to
total land area ratios in 1985 have higher growth of protected land in the periods
thereafter (two regressions show this relationship; one linear and one non-linear).
16
y = -77,571x + 23,684
R
2
= 0,3303
y = 1,3136x
-0,3942
R
2
= 0,6839
-20
-10
0
10
20
30
40
50
60
-0,01 0,04 0,09 0,14 0,19 0,24 0,29
Initial reserve quota
Fig
ure 6. Protected land accumulation growth in relation to initial reserve quota 1985.
4 Empirical analysis
This section contains empirical modeling and analysis of data. First, we briefly
investigate absolute convergence in a traditional setting (see e.g. Persson, 1997, or
Barro and Sala-i-Martin, 2004). Then we proceed to upgrade the econometric model
to include more equations, more relevant variables, and usage of the complete data set
(year-to-year). That is, conditional convergence and assessment of the determinants of
growth, where focus is put on the forest sector, tourism, and protected land.
When analyzing absolute convergence we assume that all municipalities have the
same steady-state level of the economy. Then the growth equation, as discussed in the
theory section, becomes,
17
Titit
it
Tit ybc
y
y
T+
++)(−=
ε
loglog
1,
Average annual growth between the periods t and t + T depend only on the initial
level of per capita income. In figure 7 we depict the relationship between average
annual economic growth betweeen 1985 and 2001, and the initial income level 1985.
A simple trend line (linear regression) is fitted to the data (R2 = 0.45) which shows
that, using this modeling approach, the mountain municipalities are characterized by
absolute convergence; those ecoonomies relatively poor in 1985 seem to have a larger
average annual growth rate during the period 1985-2001.
y = -0,0363x + 3,6755
R2 = 0,4532
0
0,5
1
1,5
2
2,5
45 50 55 60 65 70 75
Initial income 1985
Growth in percent
Figure 7. Absolute convergence in the Swedish mountain municipalities (base year 1985).
Next, we want to assess the role of the forest sector, the tourism sector and protected
land accumulation, and their impact on economic growth. Also, since the above
analysis of absolute convergence is limited to 15 observations (only cross-section),
18
we will now use all time periods between 1985 and 2001 (17 years times 15 cross-
sections). To analyze conditional convergence and determinants of growth in the
Swedish mountain area, we make use of the standard empirical growth model above
(see, for example, Barro and Sala-i-Martin, chapter 11, 2004) augmented with three
extra equations for variables which we assume to be also determined within the local
economy. In addition, the model is modified so that we make use of the complete data
set at hand (17 time obs and 15 regions amount to a panel of 255 obs).
In the empirical analysis below, the variables are divided into endogenous variables
and variables assumed exogenous. The endogenous variables are assumed to be
determined within the local economy, while the exogenous variables are determined
externally outside the local economy (the sub-index i denotes different municipalities,
and the sub-index t denotes different time periods). 3
Endogenous variables (all per capita):
git =
−1
log
it
it
y
y = growth in income from salary
Fit = employment in forest industry
Tit = employment in tourism sector
Nit = net migration
Exogenous variables:
3 Admittedly, some of the exogenous variables are “quasi”-exogenous, i.e., they could very
well be defined as endogenous and have their own equation. Any problems of endogeniety
are assumed to vanish as we use previous period values of these variables in the econometric
analysis.
19
GDPt = national gross domestic product per capita
Pit = protected land in % of total land
TAXit = tax rate
EDUit = above high school education (per capita)
DENSit = population density; population/total land area
PRODPOPit = part of population in ”productive” age (16-65 years)
TIMEt = linear time trend
Djt = a dummy for each county, j = NB, VB, Jämtland, Dalarna.
To simplify the notation, a vector for the exogenous variables is defined as follows;
Note that all variables in the subsequent analysis, except GDPt and TIMEt, are
municipality specific.4
Each endogenous variable is represented by an equation which include the other
endogenous variables, a vector of exogenous variables (including protected land), and
county dummys. Also, the growth equation contains a “convergence” term. The
system of equations could be estimated in reduced form, but since we are interested in
the partial effects that endogenous variables may have on each other, we have chosen
to estimate the complete system (reduced form “effects” of exogenous variables can,
however, be calculated from this system by simple substitution if necessary). The
complete model equation system is specified as follows:
4 However, this does not mean that they can not be determined to a large extent from factors
– such as national policy - outside the municipality. But the long-term levels of the variables
()
g
itjjg
j
ititFitNit
g
it eDdumTaFaNaybag +++++++= ∑
3
1=
Τ1− Xaex
log
*
[]
tititititittit TIMEPRODPOPDENSEDUTAXPGDP ,,,,,,=X
20
where,
and b is the convergence parameter.5 The parameter vectors associated with the
exogenous variable vector, Xi,t, in each equation are, aex, bex, cex, and dex respectively.
The system of equations specified above is estimated using a panel data instrumental
variable approach - Generalized method of moments (GMM).6 The covariance matrix
is robust to heteroscedasticity (in the White sense) and autocorrelation of the first
degree. The set of instrument variables is chosen so that they, at time t, are orthogonal
to the error terms at time t. In other words, the instruments are assumed to be
independent of the vector of error terms.
The estimation results are presented in Table 3 to Table 6 . We will comment on some
of the results.
can differ across regions.
5 Note that when the time period is 1 year, b* is the same as b. That is,
()
TbTb /exp
*)(−−1−= , and in this case T = 1.
6 See, for example, Bond et al (2001) and Blundell and Bond (1999) for a discussion of the
use of GMM in economic growth modelling and when estimating production functions.
N
itjjN
j
itexitTitFitg
N
it eDdumTbFbgbbN ++++++= ∑
=
3
1
'
Xb
F
itjjF
j
itexitTitNitg
F
it eDdumTcNcgccF ++++++= ∑
=
3
1
'
Xc
T
itjjT
j
itexitFitNitg
T
it eDdumFdNdgddT ++++++= ∑
=
3
1
'
Xd
()
)(−−1−= bb exp
*
21
Table 3. The growth equation (R-squared = 0.54)
Variable Estimate Stdev t-statistic P-value
Constant -.228341 .050759 -4.49851 [.000]
()
1−it
ylog -.041134 .853558E-02 -4.81911 [.000]
F .804496 .069298 11.6092 [.000]
T .050230 .090556 .554680 [.579]
N 1.46784 .186987 7.84998 [.000]
GDP 1.70708 .089383 19.0985 [.000]
P -.312095E-02 .010964 -.284652 [.776]
TAX -.214817E-02 .217369E-02 -.988260 [.323]
TIME .320319E-02 .874726E-03 3.66193 [.000]
EDU -.866470E-02 .133975 -.064674 [.948]
DENS .744036E-02 .412094E-02 1.80550 [.071]
PRODPOP .286946 .062978 4.55630 [.000]
D11 .029148 .914745E-02 3.18650 [.001]
D12 .026347 .703226E-02 3.74661 [.000]
D13 .021488 .523662E-02 4.10338 [.000]
Parameter estimates presented in table 3 suggests that there is conditional divergence
in the Swedish mountain region, which is not in line with the results on absolute
convergence presented above. Furthermore, the forest sector parameter is positive and
significant indicating that higher employment in the forest sector is good for the local
economy. The tourism parameter is also positive but not significant.7 Other variables
7 The model was modified to include tourism slope dummys for Åre, Malung, Kiruna, and
Storuman (municipalities with a relatively large tourist sector) , but none of the dummy
parameter estimates were significant.
22
that have a positive significant effect on economic growth is net migration, national
GDP, and part of population that is in productive age (16-65). Note that protected
land does not seem to have any effect on local economic growth.
Table 4. The net migration equation (R-squared = 0.27)
Variable Estimate Stdev t-statistic P-value
Constant .151954 .020734 7.32866 [.000]
g .120531 .024616 4.89654 [.000]
F -.277514 .019386 -14.3153 [.000]
T -.511488E-02 .041737 -.122550 [.902]
GDP -.251247 .046253 -5.43206 [.000]
P -.358772E-03 .334396E-02-.107290 [.915]
TAX -.620486E-03 .732013E-03-.847644 [.397]
TIME -.134254E-02 .234817E-03-5.71738 [.000]
EDU .036870 .039332 .937402 [.349]
DENS -.346385E-02 .974369E-03-3.55497 [.000]
PRODPOP -.166299 .022326 -7.44882 [.000]
D21 -.015899 .240018E-02-6.62420 [.000]
D22 -.010363 .202781E-02-5.11024 [.000]
D23 -.751622E-02 .148918E-02-5.04723 [.000]
Net migration is positively affected by local growth. That is, local economy growth
seems to attract new residents. GDP, population density, and amount of residents in
productive age are negatively correlated to net migration. This suggests that the when
the overall economy is doing good, people will move from this area of Sweden. Also,
municipalities that are relatively dense in their population, and have a relatively large
23
proportion of productive labor, will suffer larger negative net migration. The results
also suggest that protected land does not attract new residents.
Table 5. The forest sector equation (R-squared = 0.74)
Variable Estimate Stdev t-statistic P-value
Constant .319429 .053770 5.94065 [.000]
g .346381 .046103 7.51321 [.000]
T -.606085 .078530 -7.71792 [.000]
N -1.29452 .155713 -8.31354 [.000]
GDP -.517304 .112609 -4.59379 [.000]
P -.310663E-02 .755850E-02 -.411012 [.681]
TAX -.253584E-02 .204259E-02 -1.24148 [.214]
TIME -.273217E-02 .697960E-03 -3.91451 [.000]
EDU .023655 .060713 .389613 [.697]
DENS -.024820 .174675E-02 -14.2094 [.000]
PRODPOP -.157067 .040887 -3.84149 [.000]
D31 -.077451 .456924E-02 -16.9505 [.000]
D32 -.045461 .504216E-02 -9.01622 [.000]
D33 -.031912 .295390E-02 -10.8032 [.000]
The forest industry and local growth is positively correlated. The results also suggest
that if tourism goes up, forest sector activity goes down. The overall Swedish
economy, GDP, acts as a drain on forest sector employment. Interestingly, protected
land does not have any statistically significant effect on forest sector employment.
24
Table 6. The tourism equation (R-squared = 0.63)
Variable Estimate Stdev t-statistic P-value
Constant .067847 .022651 2.99528 [.003]
g -.014674 .032227 -.455342 [.649]
F -.620652 .039304 -15.7911 [.000]
N .373627 .167964 2.22445 [.026]
GDP .227976 .055946 4.07494 [.000]
P -.010096 .584239E-02 -1.72801 [.084]
TAX -.402064E-02 .987627E-03 -4.07101 [.000]
TIME .835200E-04 .568314E-03 .146961 [.883]
EDU .067047 .059089 1.13468 [.257]
DENS -.030399 .152173E-02 -19.9763 [.000]
PRODPOP .250358 .022534 11.1102 [.000]
D41 -.080803 .368366E-02 -21.9356 [.000]
D42 -.044647 .276970E-02 -16.1200 [.000]
D43 -.029382 .253109E-02 -11.6085 [.000]
Tourism is positively related to net migration. If people are moving in (out) of the
community, tourism employment goes up (down). Protected land is weakly (at 10%
level) negatively correlated with tourism employment, i.e., more protected land less
tourism.
6 Conclusions
In this paper we have studied growth patterns and determinants of growth in the
Swedish mountain municipalities from Kiruna in the north to Malung in the south.
The analysis has focused on the effects on economic growth of forest sector and
25
tourism sector employment, and the amount of protected land (in relation to total land
area). Convergence, absolute and conditional, across municipalities is also
investigated.
The results on convergence are not conclusive. Traditional growth equation modelling
estimates reveals absolute convergence (in line with neo-classical theory); that is,
using only the cross-section part of the data and assuming that all sub-regions have
the same steady-state level of their economy, the results show that initially poor
municipalities seem to grow faster than initially rich municipalities. However,
utilizing the complete data set, and assuming that the sub-regions may have different
level of steady-states, the results show divergence in per capita incomes (which would
contradict neo-classical theory).
Forest industry employment has a positive impact on local economic growth and vice
versa, while tourism seem to have no significant effect on economic growth or vice
versa. Growth of the overall economy in Sweden, measured as national GDP, seem
very important for local economy. It is positively correlated with local economic
growth and tourism, while negatively correlated with net migration, and forest sector
employment. This means that local tourism employment is determined mainly by
non-local factors. Furthermore, the results suggest that tourism employment is not a
driving force for local economic growth. A possible explanation of the GDP-forest
industry negative correlation is that the forest industry business cycle has a tendency
to develop over time counter-cyclical to the overall economy. The negative
correlation between net migration and GDP is probably due to the fact that as the
overall Swedish economy is booming, people from the mountain area move to where
they can find work; for example to cities at the coast such as Umeå, Skellefteå, and
Luleå.
26
The amount of protected land in relation to total land area does not seem to a
significant effect, at least in the period 1985-2001, on local economic growth or on
forest sector employment. This has two reasonable explanations: land protected have
been either non-productive forest land, such as mountain tops and wetlands, or that
the forest land protected was too young for harvesting, and the loss of income is not
visible in the data yet. However, future increases in the amount of protected land, that
includes productive forest land, may have a negative effect on local economic growth,
since forest sector employment is a driving force behind growth. For tourism,
however, protected land has a negative impact (at 10% statistical level). This effect
may be a result of restrictions to business operations in land areas that are protected
for some reason. These results also stand in contrast to findings in the US (see e.g.
Lorah and Southwick, 2004) where protected land have been found to attract new
residents and to be beneficial to local business activity.
In sum: forest sector employment is more important than tourism employment for
local growth; net migration is positively related to local economic growth and
negatively related to the economic growth of Sweden; protected land has had no
effect on local growth or forest sector employment, however, tourism may be
negatively affected.
An alternative modeling route would be to make use of vector auto-regressions to
assess causality between endogenous – and possibly exogenous – variables of interest
(see e.g. Berck et al, 2000). This approach would also take into account dynamics in a
more thorough and adequate manner. We leave this for future research.
References
Aghion, P. and P. Howitt (1998), Endogenous Growth Theory, MIT Press:
27
Cambridge, MA, USA.
Aronsson, T., J. Lundberg, M. Wikström (2001), Regional Income Growth and Net
Migration in Sweden, 1970 – 1995, Regional Studies 35(1), 823-30.
Barro, R. and X. Sala-i-Martin (2004), Economic Growth, MIT Press: Cambridge,
MA, USA.
Berck, P., C. Costello, L. Fortmann, S. Hoffman (2000), Poverty and Employment in
Timber-Dependent Counties, Discussion paper 00-52, Resources for the future.
Blundell, T. and S. Bond (1999), GMM Estimation with Persistent Panel Data: An
Application to Production Functions, The Institute for Fiscal Studies, Working Paper
Series No. W99/4.
Bond, S., A. Hoeffler, J. Temple (2001), GMM Estimation of Empirical Growth
Models, Economics Group Nuffield College, University of Oxford, Economic Papers
series number 2001-W21.
Eliasson, K. and O. Westerlund (2003), Vad förklarar skillnader i tillväxt mellan
svenska regioner?, Institutet för tillväxtpolitiska studier (ITPS).
Jonsson, V. (2004), Avverkning, naturreservat och dess betydelse för ekonomisk
tillväxt i åtta fjällkommuner 1985-2001, FjällMistra-rapport 6, SLU, Umeå, Sweden.
Also published as a master thesis at the department of economics, university of Umeå,
Sweden.
Lorah, P., Southwick, R. (2003), Environmental Protection, Population Change, and
28
Economic Development in the Rural Western United States, Population and
Environment 24(3), 255-71.
Lundberg, J. (2001), On the Determinants of Average Income Growth and Net
Migration at the Municipal Level in Sweden, Working paper 34:2001, CERUM.
Persson, J. (1997), Convergence Across the Swedish Counties, 1911-1993, European
Economic Review 41, 1835-52.
Rasker, R. and B. Alexander (2003), Working Around the White Clouds, Sonoran
Institute, Tuscon, AZ.
Skontoft, A. and H. Solem (2001), Economic Growth and Land-Use Changes: the
Declining Amount of Wilderness Land in Norway, Ecological Economics 37, 289-
301.
Solow, R. M. (1956), A Contribution to the Theory of Economic Growth, Quartely
Journal of Economics 70:1, 65-94.
Swan, T.W. (1956), Economic Growth and Capital Accumulation, Economic Record
32, 334-361.
29
Appendix 1
The tourism employment data is calculated by counting the amount of people
employed in the following sectors:
SNI92 Keys for tourism employment data
55111 Hotels with restaurant, except conference centres
55112 Conference centres, with lodging
55120 Hotels and motels without restaurant
55210 Youth hostels, etc.
55220 Camping sites etc. incl. Caravan sites
55230 Other short-stay lodging facilities
55300 Restaurants
55400 Bars
55521 Catering for the transport sector
61200 Inland water transport
62100 Scheduled air transport
62200 Non-scheduled air transport
63210 Other supporting land transport activities
63301 Activities of tour operators
63302 Activities of travel agencies
63303 Tourist assistance
92320 Operation of arts facilities
92330 Fair and amusement park activities
92340 Other entertainment activities
92520 Museum activities and preservation of historical sites
92530 Botanical and zoological gardens and nature reserves activities
92611 Operation of ski facilities
92612 Operation of golf courses
30
92722 Operation of recreational fishing waters
92729 Various other recreational activities
52485 Retail sales of sports and leisure goods
5021 Fish farming
