For Peer Review Only: Individual Wages and Regional Market Accessibility

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 28

Spatial Economic Analysis

Fo
rP

Individual wages and regional market accessibility


ee

Journal: Spatial Economic Analysis

Manuscript ID RSEA-2020-0169
rR

Manuscript Type: Manuscript

Market Potential, Accessibility, Regional Price Deflators, Regional Labour


ev

Keywords:
Markets

J30 - General < J3 - Wages, Compensation, and Labor Costs < J - Labor
JEL Codes: and Demographic Economics, J60 - General < J6 - Mobility,
iew

Unemployment, and Vacancies < J - Labor and Demographic Economics

This paper reconsiders the relationship between individual wages and


regional market access. We focus on Poland – a country that has
experienced profound changes in market accessibility over the last two
decades, first as a consequence of EU accession in 2004, and second as
On

Abstract: a result of the unprecedented development of the major road transport


network. Using data from the years 2000-2015 we find that the positive
relationship between wages and regional market access disappears after
2004. At the same time, individual wages begin to be positively
ly

correlated with the market potential of Poland’s main trading partners.

URL: http://mc.manuscriptcentral.com/rsea
Page 1 of 27 Spatial Economic Analysis

1
2
3
4
Individual wages and regional market accessibility
5
6
7
8
9
10 Abstract
11
12
13 This paper reconsiders the relationship between individual wages and regional market access.
14
15 We focus on Poland – a country that has experienced profound changes in market accessibility
16
17 over the last two decades, first as a consequence of EU accession in 2004, and second as a result
18
Fo

19
of the unprecedented development of the major road transport network. Using data from the
20 years 2000-2015 we find that the positive relationship between wages and regional market
21
rP

22 access disappears after 2004. At the same time, individual wages begin to be positively
23
24 correlated with the market potential of Poland’s main trading partners.
ee

25
26
27 Keywords: Market Potential; Accessibility; Regional Price Deflators; Regional Labour
28
rR

29 Markets.
30
31 JEL classification: C26; J30; J60.
ev

32
33
34
iew

35
36
37
38
39
40
On

41
42
43
44
ly

45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

1
URL: http://mc.manuscriptcentral.com/rsea
Spatial Economic Analysis Page 2 of 27

1
2
3 1. Introduction
4
5
6 In new economic geography (NEG) models, regional price differentials play a key role in
7
determining the spatial allocation of economic activity. The “price index effect” constitutes one
8
9 of the most important agglomeration forces that makes both workers and firms cluster together.
10
11 Workers close to bigger markets receive both higher nominal and real wages; firms get access
12
13 to cheaper intermediate goods. As a consequence, the NEG models can be considered as an
14 approach providing micro-economic foundations for the empirical measures of market access
15
16 usually called market potential (e.g. Redding, 2013).
17
18
Fo

19 Introduced first by Harris (1954), the notion of market potential and its relationship with
20
21 regional wages has been the subject of many empirical studies. Yet, due to the lack of reliable
rP

22
data on regional price indices, there are no studies that directly include the latter in measures of
23
24 market potential. Instead different approaches are applied in the literature to deal with this
ee

25
26 problem. The first one relies on a proxy variable and assumes implicitly that price indices are
27
28 equal across regions (e.g. Da Mata et al., 2007; Mion and Naticchioni, 2005). Another variant
rR

29 by Hanson (2005) proposes measuring market potential assuming equalisation of real wages
30
31 across regions. Finally, Redding and Venables (2004) try to capture regional price indices
ev

32
33 through fixed effects.
34
iew

35
36 This paper attempts to reassess the relationship between individual wages and accessibility to
37
regional markets, taking as an example Poland – a country that has experienced profound
38
39 changes in accessibility over the last two decades, first as a consequence of accession to the EU
40
On

41 in 2004 and, second, as a result of the unprecedented development of the major road transport
42
43 network. EU accession in 2004 has removed all remaining trade barriers that existed between
44 Poland and the EU, in particular in the case of agricultural goods. However it also allowed
ly

45
46 Poland to acquire access to the EU’s structural funds. To a great extent, these were used to
47
48 finance public infrastructure investment. In particular, between 2005 and 2015 Poland invested
49
around EUR 24 billion in major road infrastructure projects of which over 50% was financed
50
51 by the EU. As a result, the overall length of high-speed road network has increased from less
52
53 than 500 km to more than 3000 km, significantly changing the overall accessibility of individual
54
55 regions.1
56
57
58
59
60 1 For more details see Komornicki et al. (2013).

2
URL: http://mc.manuscriptcentral.com/rsea
Page 3 of 27 Spatial Economic Analysis

1
2
3 In order to verify whether the accessibility change influenced the relationship between nominal
4
5 wages and regional market access we employ two different methodological approaches. Firstly,
6
7 we estimate the regional market potential derived from the Krugman (1992) model and its
8
relationship with regional wages by using actual regional price indices. Hence, we do not have
9
10 to rely on the unrealistic assumptions used in previous empirical studies. In this sense, this is
11
12 the first empirical paper that, to the best of our knowledge, directly follows the assumptions of
13
14 the NEG models. Also, in contrast to previous empirical studies, we use the actual travel time
15 indicator to measure the significance of distance which allows us to account for regional
16
17 accessibility improvements. Secondly, we validate the adequacy of an alternative measure of
18
Fo

19 regional market access that is the potential accessibility indicator frequently used in
20
21
transportation literature (e.g. Geurs and van Wee, 2004). To the best of our knowledge, such a
rP

22 measure has not previously been applied as a regional wage determinant. In addition, we study
23
24 the effect of both measures on individual wages having controlled for the market potential of
ee

25
26 Poland’s main foreign trading partners.
27
28
rR

29 The remainder of the paper is organised as follows. The next section presents a literature review
30 related to regional market access and its relationship with regional wages. Section 3 discusses
31
ev

32 the research methodology and the data used in our analysis. In section 4, we present empirical
33
34 results for the Polish regions. Finally, section 5 presents conclusions.
iew

35
36
37 2. Literature review: Market access and wages
38
39
40
Neoclassical theory suggests that under certain conditions, wages would be equalised across
On

41 regions. This, however, is hardly ever observed in the real world and as a consequence several
42
43 competing theoretical approaches were developed in order to explain the existence of regional
44
ly

45 wage disparities. These include a modified neoclassical model by Harris and Todaro (1970),
46
the core-periphery model within the NEG framework pioneered by Krugman (1991a, b; 1992)
47
48 and the wage curve approach proposed by Blanchflower and Oswald (1990). Still, empirical
49
50 literature mostly focuses on verifying only two of these: the NEG and the wage curve approach.
51
52
53 The NEG models underline the importance of market access as a determinant of regional wages.
54
55
However, even though the importance of market access was highlighted in the preceding
56 theoretical approaches (e.g. location theory), the NEG framework provides micro-economic
57
58 foundations for empirical measures of market access, usually called the regional market
59
60 potential (RMP). The regional market potential, already introduced by Harris (1954), shows

3
URL: http://mc.manuscriptcentral.com/rsea
Spatial Economic Analysis Page 4 of 27

1
2
3 that the demand for goods produced in a given region is the sum of the purchasing power in
4
5 other regions, weighted by trade costs (e.g. Hanson, 2005). In accordance with the relationship
6
7 between wages and the RMP, the former should be higher both in nominal and real terms at the
8
economic core and lower at the periphery. This is due to the stronger linkages between firms
9
10 and consumers that result in higher nominal wages, as well as lower trade costs within the
11
12 bigger market that should result in lower prices of manufacturing goods.
13
14
15 A large number of empirical studies have confirmed the positive relationship between the RMP
16
17
and regional wages. Two main differences can be distinguished in terms of the empirical
18
Fo

strategy applied. First, different definitions of market potential and different estimation
19
20 techniques are used to verify the impact of market access on wages. This includes the Harris
21
rP

22 measure of market potential (e.g. Hanson, 2005; Head and Mayer, 2006 or Hering and Poncet,
23 2010) and the market potential function derived from the Krugman model. The latter is either
24
ee

25 estimated by applying a one-step non-linear estimation approach (e.g. Brakman et al., 2004;
26
27 Hanson, 2005) or by a two-step approach based on the gravity equation (e.g. Head and Mayer,
28
rR

29 2006; Lovely et al., 2019). Second, the existing studies differ with regard to the wage data. The
30 majority of the existing papers apply aggregated regional data. The examples of such studies
31
ev

32 include papers by Feenstra and Hanson (2000) and Hanson (2005) on US counties, Mion (2004)
33
34 for Italy, Brakman et al. (2004) for German districts and Head and Mayer (2006) and Breinlich
iew

35
(2006) for the European Union. Relatively few papers rely on microdata that allow worker and
36
37 firm characteristics to be controlled (e.g. Fally et al., 2010 for Brazil; Hering and Poncet, 2010
38
39 for China, Matas et al. 2015 for Spain, and Cieślik and Rokicki, 2016 for Poland). The latter
40
On

41 group allows important elements of labour economics, which are neglected in the NEG wage
42 equations, to be taken into account. Furthermore, it can alleviate the potential problem of
43
44 endogeneity associated with market access since it is less likely that a shock to an individual
ly

45
46 wage translates into a change in local market access.
47
48
49 In general, there is no significant difference in results between papers based on aggregate
50
regional data and papers that apply individual microdata. The estimated coefficient on the
51
52 regional market potential is positive and in the majority of cases ranges between 0.05 and 0.30.
53
54 Higher estimates are usually obtained in the regressions that employ the Harris measure of
55
56 market potential as compared to the Krugman model. In the case of the former the estimated
57 market potential coefficient can even reach 0.7 (e.g. Hering and Poncet, 2010). Still, the actual
58
59 estimates of the relationship between the regional market potential and regional wages are
60

4
URL: http://mc.manuscriptcentral.com/rsea
Page 5 of 27 Spatial Economic Analysis

1
2
3 subject to a possible bias due to the fact that the RMP is either calculated upon unrealistic
4
5 assumptions, such as regional price or regional real wage equalisation, or is estimated through
6
7 the gravity equation. The main reason for this is the lack of appropriate data, hence none of the
8
existing empirical studies was able to directly employ regional price deflators as a variable used
9
10 to calculate the RMP.
11
12
13 Moreover, there are only a very few studies on regional wage determinants that take into
14
15 account improvements in transportation infrastructure that affect regional accessibility. This is
16
17
surprising given the fact that the importance of a transportation network as a determinant of
18
Fo

location decisions has been underlined in many papers (e.g. Alañón et al., 2007; Combes et al.,
19
20 2008; Jiang et al., 2018). The notable exception is the study by Matas et al. (2015) who
21
rP

22 estimated the wage equation with individual data for Spanish workers in order to provide
23 evidence on the effect of improvements in road transport infrastructure. However, their wage
24
ee

25 equation was not derived directly from any formal theoretical model.
26
27
28 Therefore, the contribution of this paper to the literature is threefold. First, this is the first paper
rR

29
30 that uses actual regional price deflators to estimate the RMP and its relationship with individual
31
wages. This allows us to: i) validate whether the use of regional price deflators significantly
ev

32
33 affects the RMP and their spatial pattern and ii) re-estimate the wage equation and see what is
34
iew

35 the “true” RMP coefficient. Second, in contrast to the previous empirical studies we use the
36
37 actual travel time indicator that allows us to account for regional accessibility improvements.
38 The use of the latter is particularly important in the case of a country that has experienced
39
40 radical improvements in the transportation infrastructure over recent years. Third, we verify
On

41
42 the adequacy of an alternative measure of market access that is the potential accessibility
43
44
indicator. This measure is very frequently used in the transportation literature. However, it has
ly

45 never been employed in empirical research on wage determinants. This allows us to compare
46
47 the effects of the measures used in both economic geography and transportation literature on
48
49 individual wages.
50
51
52 3. Research methodology and statistical data
53
54
55
3.1. Research methodology
56
57 In this paper we compare the results of empirical studies based on two different approaches.
58
59 The first approach validates the relationship between nominal wages and the regional market
60

5
URL: http://mc.manuscriptcentral.com/rsea
Spatial Economic Analysis Page 6 of 27

1
2
3 potential derived from the model by Krugman (1992). In the second empirical approach, instead
4
5 of using the Krugman’s market potential model, we employ the potential accessibility indicator
6
7 commonly applied in the transport geography literature.
8
9
10 Krugman’s market potential model assumes the existence of a two-sector economy consisting
11 of agriculture (A) and manufacturing (M). The A-sector is characterised by constant returns to
12
13 scale (CRS), perfect competition and no trade costs between regions. The M-sector firms
14
15 operate under increasing returns to scale (IRS) and produce differentiated products. Each
16
17
variety is produced by a single firm under monopolistic competition and there are “iceberg”
18
Fo

transportation costs between R regions.2 Consumer preferences are described by the standard
19
20 Cobb-Douglas utility function with a Dixit-Stiglitz sub-utility for M-good and assumed to be
21
rP

22 the same across regions:


23
24
U  CM C1A  ; 0 < μ < 1 (1)
ee

25
26 where:
27
28
rR

29  1

30  N   1
CM    q (v )  dv  (2)
31  0

 
ev

32
33
34
and CM is an index of the consumption of the M-sector varieties, which are defined as a
iew

35
36 continuum of N goods, where q(v) corresponds to the demand for the v-th variety, σ denotes the
37
38 constant elasticity of substitution (CES) between any two varieties (σ > 1), while μ denotes the
39
40 fixed proportion of income that is devoted to consumption of varieties of M-good.
On

41
42
43 The consumer maximises utility function (1) with respect to the usual budget constraint to
44
ly

obtain region s’s demand for a variety produced in region r:


45
46
47 
prs
48 qrs(v) = Ys (3)
49 Gs1
50
51
52 where μYs is region s’s total expenditure on M-good and prs is the consumer price of a variety
53
54 from region r sold in region s.3 Demand qrs(v) is determined by regional income Ys, consumer
55
56
57
58
59
2The multi-region model was first proposed by Krugman (1992) and was later summarised in Fujita et al. (1999).
3The consumer price in region s includes factory price pr and iceberg transportation costs τrs between regions r
60
and s, hence prs = prτrs.

6
URL: http://mc.manuscriptcentral.com/rsea
Page 7 of 27 Spatial Economic Analysis

1
2
3 price prs and the regional price index for M-good Gs expressed as:
4
5
6 1
7  1  
1 

8 Gs =  nr prs  (4)
9
 r 
10
11
12
where nr is the number of varieties produced in region r.
13
14
The total cost of producing qr in region r is assumed to take the form: crqr + Fr, where Fr is a
15
16 plant-specific fixed cost, and cr is the constant marginal cost of production. There is also the
17
18 iceberg transportation cost τrs, between any two regions r and s. Hence, the gross profit function
Fo

19
20 of a firm located in region r on market s is given by:
21
rP

22
Πrs = (pr − cr)τrsqrs (5)
23
24
ee

25 The solution to the firm’s profit maximisation problem yields the optimal factory price for each
26
27 origin r as a constant mark-up over marginal costs:
28
rR

29
30 cr
31
pr = (6)
 1
ev

32
33
34 As a result all varieties produced in region r have the same factory price and the gross profit
iew

35
36 earned in market s for a variety produced in region r is given by Πrs = prxrs/σ.
37
38
39 The total profits earned by a firm located in region r are calculated by substituting factory price
40
On

41
(6) into equation (5), summing up profits over all regions s, and substracting fixed cost Fr:
42
43 R
 R 
Πr =  pr xrs /  − Fr = c1r    rs Gs Ys  − Fr
44  1
ly

(7)
45 s 1  s 1 
46
47
48 1
49 where: λ =     and Φrs = τrs1-σ represents the total impact of transportation cost τrs,
50    1
51
52 and the elasticity of substitution σ on the demand which is interpreted as the freeness of trade
53 and takes values between 0 and 1.4
54
55
56 A part of profit function (7) can be interpreted as the regional market potential of region r
57
58
59 4When Φrj = 0 trade costs become prohibitive and only the local demand is relevant. Alternatively, in a frictionless
60
world when Φrj = 1 trade costs are equal zero.

7
URL: http://mc.manuscriptcentral.com/rsea
Spatial Economic Analysis Page 8 of 27

1
2
3 (RMPr). It shows how well location r is endowed with access to markets for the varieties of M-
4
5 good that it produces:
6
7
8  R 
RMPr =   rs Gs Ys 
 1
9 (8)
10  s 1 
11
12
13 The equilibrium in the model occurs when all firms earn the same level of profit which is equal
14
to zero in the long run. This yields the relationship between the RMP and costs:
15
16
17 λRMPr = Fr cr 1 (9)
18
Fo

19
20 Equation (9) implicitly determines the relationship between the wage rate and the RMP.
21
rP

22 However, derivation of the direct relationship between the wage rate and the RMP requires
23
24 additional assumptions regarding the production technology of the M-good. Following
ee

25
26
Krugman (1991a, 1992), we assume that labour is the sole factor of production and a labour
27 requirement can be defined for fixed cost Fr and for marginal cost cr. In this case the region-
28
rR

29 specific fixed labour requirement is denoted by αr and the constant labour requirement per unit
30
31 of output by βr.
ev

32
33
34 Rearranging the relationship between the RMP and the costs given by equation (9) and
iew

35 replacing Fr and cr by their respective expressions yields the following relationship between
36
37 wages and the RMP:
38
39
40 1
On

41   
42 wr =  RMPr  (10)
 r  r
 1
43 
44
ly

45
46 Equation (10) is the standard wage equation that indicates the nominal wage rate an individual
47 firm from location r is willing to pay a worker in this location. It implies that firms in locations
48
49 with a higher RMP incur smaller transportation costs and offer higher wages.
50
51
52 In deriving the aforementioned wage equation it was assumed that firms and workers in the M-
53
54 sector were exactly the same. This assumption is, however, rejected by the literature on labour
55 economics. This literature stressed the importance of the characteristics of both worker and firm
56
57 in determining individual wage rates. Therefore, this unrealistic assumption needs to be relaxed.
58
59 Following the approach used earlier by Fally et al. (2010), Hering and Poncet (2010) and
60

8
URL: http://mc.manuscriptcentral.com/rsea
Page 9 of 27 Spatial Economic Analysis

1
2
3 Cieślik and Rokicki (2016), we assume that in each region r individual labour requirement lir
4
5 depends both on output per region and the set of worker and firm characteristics zir:
6
7
8 lir = [αr + βrqr] exp(−ρzir), (11)
9
10
11 This approach allows the relationship between individual wages and the RMP to be studied,
12
13
having controlled for the set of worker and firm characteristics:
14
15 1
16   
17 wir =  RMPr exp(ρzir) (12)
 r  r
 1
18 
Fo

19
20
21 Log-linearising (12) yields an empirically testable relationship between individual wages, the
rP

22
23 RMP and the set of individual worker and firm characteristics of the following form:
24
ee

25
26 logwijrt = θ logRMPrt + ρzijrt + γxijrt + μr + δj + νt + εijrt (13)
27
28
rR

29
where: wijrt is the nominal wage rate of an individual worker i employed in industry j at location
30 r at time t, RMPrt is the value of market potential of region r at time t and is given by the
31
ev

32 equation (8), zijrt is the set of characteristics of an individual worker i employed in industry j in
33
34 location r at time t, xijrt is the set of characteristics of a firm and city in which an individual
iew

35
worker i is employed at time t, μr is the region specific effect, δj is the industry specific effect,
36
37 νt is the time specific effect, εijrt is the error term. The parameter θ on the log of the RMP is the
38
39 elasticity of the RMP in relation to the individual wage, ρ is the estimated vector of parameters
40
On

41 on the set of individual worker characteristics, and γ is the estimated vector of parameters on
42 firm and city characteristics.
43
44
ly

45 In our second empirical approach we replace the RMP variable in equation (13) by the potential
46
47 accessibility indicator frequently used in the transportation literature (e.g. Geurs and van Wee,
48
49 2004). The measure is based on two assumptions: i) larger centres are more important than
50 smaller ones and ii) the attractiveness of destinations decreases with distance (Hansen, 1959).
51
52 The attractiveness of a given destination is measured by its population. On the other hand, travel
53
54 time is used as an impedance component and a negative exponential within the distance decay
55
56
function (e.g. Geurs and Ritsema van Eck, 2001). The potential accessibility indicator is given
57 by:
58
59
60

9
URL: http://mc.manuscriptcentral.com/rsea
Spatial Economic Analysis Page 10 of 27

1
2
3 𝐴𝐶𝑟 = ∑𝑠𝑀𝑠𝑒𝑥𝑝( ―𝛽𝑡𝑟𝑠) (14)
4
5
6
7
where ACr is the accessibility of location r, Ms represents attractiveness of destination j (i.e.
8 population) and trs is a travel time between r and s. Following the approach of Rosik et al.
9
10 (2015) the parameter β has a constant value of 0.013862. Our potential accessibility indicator
11
12 is based on origin-destination matrices of travel times between locations within the Polish
13 borders. In accordance with the ‘half-life’ approach (Östh et al. 2014), the β parameter
14
15 corresponds to a typical trip duration of 50 minutes. With the inclusion of the potential
16
17 accessibility indicator in the equation for our estimate takes the form of:
18
Fo

19
20 logwijrt = θ logACrt + ρzijrt + γxijrt + μr + δj + νt + εijrt (15)
21
rP

22
23 where ACrt is the potential accessibility of location r at time t, while other variables are exactly
24 the same as in equation (13).
ee

25
26
27 Note, that in the case of both empirical approaches we also test specifications where apart from
28
rR

29 the regional market potential we also added the external market potential. The latter relies on
30
31 the GDPs of Poland’s main trade partners discounted by distance between the capital cities.
ev

32
This variable is included due to the fact that the spatial distribution of average wages in Poland
33
34 is clearly skewed towards the European core. Figure 1 shows the spatial distribution of gross
iew

35
36 monthly wages in 2015. It can easily be noticed that the majority of the regions with the highest
37
38 relative level of wages are situated along Poland’s western and southern borders. The notable
39 exception is the Mazowieckie, the biggest voivodeship, with the capital city - Warsaw.
40
On

41
42 <<Insert figure 1 here>>
43
44
ly

45
46 3.2. Statistical data
47
48
In this paper we use both micro and macro-level data. The micro-level wage related data were
49
50 obtained from the Polish Labour Force Survey (LFS) at the 16 NUTS2 level regions over the
51
52 period 2000 - 2015. The survey did not gather information on wages of the self-employed or
53
54 paid family workers and was composed of approximately 40 thousand observations per wave5.
55 After excluding the unemployed, inactive and missing observations, we are left with 142,112
56
57 observations of which 74,050 relate to males and 68,062 to females. The individual level data
58
59
60 5 The LFS survey is conducted quarterly. Our dataset is based on the 4th quarter wave from each year.

10
URL: http://mc.manuscriptcentral.com/rsea
Page 11 of 27 Spatial Economic Analysis

1
2
3 set allows us to control for a large set of individual characteristics affecting individuals’ wage
4
5 responses to regional variations in market access. Data on firm characteristics enable us to also
6
7 control for possible agglomeration effects.6
8
9
10 In accordance with equation (8), the RMP requires information on regional incomes, regional
11 price indices and distance measures. In our study we use the regional GDP, initially expressed
12
13 in millions of Polish zloties (PLN) at current market prices and then converted into 2010
14
15 constant prices, as a measure of regional income. The data comes from the Polish Regional
16
17
Statistics database (Bank Danych Lokalnych) compiled by the Polish CSO and available online
18
Fo

at http://www.stat.gov.pl/bdl. The data on regional price indices for the Polish NUTS2 level
19
20 regions is calculated in accordance with the approach used by Rokicki and Hewings (2019).
21
rP

22 Finally, we tested three different distance measures. The first one defines the distances between
23 the capitals of different regions as a travel time (passenger car) in 2015 expressed in minutes.
24
ee

25 The travel time data is retrieved from the ViaMichelin route planner available online (at
26
27 http://www.viamichelin.pl/). Our second and third measures rely on travel times calculated
28
rR

29 using the distance along the road network and a truck speed model – hence we call it the
30 accessibility based measure. The model is based on 14 pre-defined categories of roads in Poland
31
ev

32 and official maximum speeds derived from the Polish Highway Code, adjusted down for driving
33
34 impediments, including the presence of built-up areas, topography and population density.7 We
iew

35
calculate travel times between nodes that represent all municipalities (LAU2 units), receiving
36
37 accessibility values at the municipal level which are then aggregated to regional NUTS2 units.
38
39 The difference between the second and the third measure is that the second assumes a constant
40
On

41 2015 accessibility level while the third accounts for accessibility improvements over the years
42 2000-2015. Travel times based on the truck speed model are also used in calculating the
43
44 potential accessibility indicator. In this case we use population weighted averages. Population
ly

45
46 data comes from the Polish Statistical Office.
47
48
49 The trading partners’ external market potential is measured in three different ways. First, we
50
calculate the potential of the main trading partners weighted by their share of Polish exports.
51
52 Poland’s main trading partners include: Austria, Belgium, the Czech Republic, Denmark,
53
54 France, Germany, Hungary, Italy, Netherlands, Slovakia, Spain, Sweden and the UK (all
55
56
57
58
59 6 See the Appendix for a detailed description of the dataset.
60 7 For details see Śleszyński (2014).

11
URL: http://mc.manuscriptcentral.com/rsea
Spatial Economic Analysis Page 12 of 27

1
2
3 partners that account for more than a 2% share of Polish exports).8 The trading partners external
4
5 market potential is measured as (version 1):
6
7
8 𝑇𝑀𝑃𝑟𝑡 = ∑𝑗𝐺𝐷𝑃𝑗𝑡*weightjt/distance between capitalsjr (16)
9
10 Where the weight is the share of exports to each partner in total Polish exports in each year
11
12 (row standardised). By the distance between capitals we refer to the distance between region r
13 and capital of the partner country.
14
15
16 In addition, we also calculate two simpler indicators that may affect regional wages in Poland.
17
18 The second indicator is calculated as the sum of trading partners’ GDPs discounted by distance
Fo

19
20 from each regional capital to Brussels. The trading partner’s market potential is calculated as
21
rP

(version 2):
22
23
24 𝑇𝑀𝑃𝑟𝑡 = (∑𝑗𝐺𝐷𝑃𝑗𝑡)/distance to Brusselsr (17)
ee

25
26
where: TMPrt denotes the trading partner’s market potential for region r at time t, GDPjt denotes
27
28 the GDP of country j at time t.
rR

29
30
31 As a special case we focus on Germany as Poland’s most important trading partner. In this case
ev

32
33 the trading partner’s potential for Germany is measured as (version 3):
34
iew

35
36 𝑇𝑀𝑃𝑟𝑡 = 𝐺𝑒𝑟𝑚𝑎𝑛𝐺𝐷𝑃𝑡/distance to Hanoverr (18)
37
38
39
40
On

where Hanover is chosen as a city placed in the geographic centre of Germany.


41
42
43 The data on trading partners’ GDPs comes from EUROSTAT and is expressed in million euro
44
ly

45 at constant 2010 prices. The data is then converted into million PLN using the average exchange
46
47 rate provided by National Bank of Poland. Travel times between regional capitals and different
48
European cities come from the ViaMichelin route planner available online (at
49
50 http://www.viamichelin.pl/).
51
52
53 4. Empirical results
54
55
56 In this section we report and discuss our empirical results. First, in Table 1 we show the
57
58
59 8Over the period analysed the overall share of the aforementioned countries in Polish exports varied between
60
67.4% (2013) and 72.9% (2009 and 2010).

12
URL: http://mc.manuscriptcentral.com/rsea
Page 13 of 27 Spatial Economic Analysis

1
2
3 benchmark estimation results obtained using the Krugman approach for the 2000-2015 period
4
5 with three different distance measures, having controlled for individual worker and firm
6
7 characteristics, and regional and trade indicator variables but without controlling for the
8
external market potential of Poland’s major trading partners. In other words, we initially focus
9
10 entirely on the role of different distance measures in the estimation of Poland’s internal market
11
12 potential and disregard the role of the external market potential. All regressions based on the
13
14 Krugman approach are estimated by nonlinear least squares (NLS) while the regressions based
15 on the potential accessibility indicator approach are estimated by ordinary least squares (OLS).
16
17
18
Fo

<<Insert table 1 here>>


19
20
21 In column (1) of Table 1 we report the results of estimates obtained using 2015 travel time
rP

22
23 between the Polish regional capitals. It turns out that the estimated parameter of the regional
24
market potential is already statistically significant at the 1% level of statistical significance and
ee

25
26 displays the expected positive sign. The estimated parameter of the distance variable is also
27
28 statistically significant at the 1% level but shows a positive sign which is not in line with the
rR

29
30 theory’s predictions. Finally, the estimated parameter of the RPP deflator is also significant at
31
the 1% level and displays a positive sign which is in line with expectations. In column (2) we
ev

32
33 show the results of estimates obtained using a constant 2015 measure of accessibility. These
34
iew

35 results look similar to the results reported in column (1) but the estimated coefficient of the
36
37 distance variable is not now statistically significant. Finally, in column (3) we report the results
38 of estimates obtained from the specification in which we accounted for accessibility
39
40 improvements. In this case it can be seen that the sign of the estimated parameter of the distance
On

41
42 variable is correct but it remains not statistically significant. Furthermore, the magnitude of the
43
44
estimated parameter on the internal market potential variable (0.079) is now significantly lower
ly

45 compared to previous specifications (both around 0.2) and its statistical significance drops to
46
47 only 5%. This may suggest that once we take into account accessibility improvements, the
48
49 internal market potential in Poland becomes less important for wage determination. It also
50
indicates that previous studies based on constant distance measures are likely to overestimate
51
52 the impact of market potential on nominal wages.
53
54
55 In Table 2 we report the results of estimates obtained from the specification in which we account
56
57 for accessibility improvements for different sub-periods of our sample to study the effects of
58
59
the European integration. For the sake of easier comparison, in column (1) we show once again
60 the results of estimates already reported in column (3) of Table 1 obtained for the entire sample,

13
URL: http://mc.manuscriptcentral.com/rsea
Spatial Economic Analysis Page 14 of 27

1
2
3 while those in columns (2)-(5) are for different sub-periods. It turns out that in contrast to the
4
5 results of estimates obtained for the whole sample, the market potential in all the sub-samples
6
7 is not statistically significant. Moreover, neither the distance variable nor the RPP deflator
8
variable are significant as well.
9
10
11 <<Insert table 2 here>>
12
13
14 In Table 3 we report results of the estimates obtained from the specification in which we account
15
16 for accessibility improvements, having controlled for the external market potential (potential of
17
18 Poland’s major trading partners) calculated using formula (16). The particular columns in Table
Fo

19 3 are the direct counterparts of the columns from Table 2. In column (1) we report results of the
20
21 estimates obtained for the entire period 2000-2015. It turns out that these results are very similar
rP

22
23 to the results reported in column (1) in Table 2 as the estimated coefficient of the external
24
market potential variable is not statistically significant. However, in the post accession period
ee

25
26 this variable becomes statistically significant at the 5% level with values over 1.2 while the
27
28 internal market potential completely loses its previous statistical significance. Therefore, it can
rR

29
30 be argued that in more recent years access to the Poland’s main export partner markets in the
31
EU plays a much more important role in wage determination than access to the Polish market.
ev

32
33 This finding clearly reflects increasing integration between Poland and its main trading partners
34
iew

35 from the EU over time.


36
37
38 <<Insert table 3 here>>
39
40
On

41 In Table 4 we report results of the estimates obtained from the alternative approach based on
42
the potential accessibility indicator without controlling for the external market potential. It turns
43
44 out that the conclusions obtained on the basis of this atheoretical approach are quite similar to
ly

45
46 the results obtained on the basis of the formal Krugman model reported in Table 2. In other
47
48 words the accessibility to internal markets in Poland is positively correlated with nominal wages
49 over the entire 2000-2015 period. Moreover, the value of the coefficient is very similar (0.105
50
51 versus 0.079). However, it loses its statistical significance after Poland’s accession to the EU.
52
53
54 <<Insert table 4 here>>
55
56
57 In Table 5 we show results of the estimates obtained from the potential accessibility indicator
58
59
approach having controlled for the external market potential. Once again, it turns out that the
60 conclusions obtained on the basis of this approach are very similar to the results obtained on
14
URL: http://mc.manuscriptcentral.com/rsea
Page 15 of 27 Spatial Economic Analysis

1
2
3 the basis of the formal Krugman model reported in Table 3. In other words, after Poland’s
4
5 accession to the EU the estimated parameter of the external market potential becomes
6
7 statistically significant. The values of the coefficient are similar to the ones observed in the
8
regressions based on the Krugman approach (around 1.1).
9
10
11 <<Insert table 5 here>>
12
13
14 In Table 6 we report results of the estimates obtained from the specification in which we used
15
16 the regional accessibility measure and we controlled for both external market potential and for
17
18 accessibility spillover. It turns out that the statistical significance of the accessibility spillover
Fo

19 variable disappears after Poland’s accession to the EU. Interestingly, in contrast to the previous
20
21 results, the estimated parameter of the external potential variable is now statistically significant
rP

22
23 both before and after Poland’s accession to the EU.
24
ee

25
26 <<Insert table 6 here>>
27
28
rR

29 To verify the robustness of our empirical results, in Table 7 we report results of the estimates
30 obtained using the instrumental variables approach. Given the fact that the results of both the
31
ev

32 Krugman model based approach and the potential accessibility measure approach are very
33
34 similar, we study only the latter. Following other studies (e.g. Matas et al., 2015), we instrument
iew

35
36
the potential accessibility indicator by its first lag. The results confirm our previous findings.
37 In other words, accessibility to internal markets in Poland becomes statistically insignificant
38
39 after EU accession. At the same time the external market potential variable remains positive
40
On

41 and statistically significant both before and after Poland’s accession to the EU. Still, the results
42
of the endogeneity test suggest that the potential accessibility variable should be treated as
43
44 exogenous. As a consequence, the previously discussed OLS results should be considered as
ly

45
46 the preferred ones.
47
48
49 <<Insert table 7 here>>
50
51
52 In the Appendix we show additional robustness tests where the external market potential was
53 calculated using simpler formulas (17) and (18). The previous conclusions still hold. The
54
55 estimated coefficients on all our measures of external market potential remain statistically
56
57 significant in all estimated specifications. In particular, our estimated results show that access
58
59
to the market of Poland’s most important trading partner – Germany contributes significantly
60 to the explanation of the variation of the individual wages of Polish workers.
15
URL: http://mc.manuscriptcentral.com/rsea
Spatial Economic Analysis Page 16 of 27

1
2
3 5. Conclusions
4
5
6 In this paper we reconsidered the empirical relationship between individual wages and
7
8 accessibility to regional markets in Poland using data for the years 2000-2015. First, we
9
10 employed region-specific price deflators to estimate the effect of the regional market potential
11 using the wage equation derived directly from the multi-region model by Krugman (1992).
12
13 Unlike previous empirical studies, in order to measure the significance of distance we used the
14
15 actual travel time indicator that allowed us to account for regional accessibility improvements.
16
17
Second, we assessed the adequacy of an alternative measure of regional market access that was
18
Fo

the potential accessibility indicator. In addition, we studied the effect of both measures on
19
20 individual wages having controlled for the market potential of Poland’s main foreign trading
21
rP

22 partners.
23
24
We showed that the application of different distance measures significantly influences
ee

25
26 estimated results based on the Krugman model. In particular, the inclusion of the time-variant
27
28 variable accounting for accessibility improvements led to a drastic drop in the regional market
rR

29
30 potential coefficient (0.08 versus 0.2). This may suggest that previous studies based on constant
31
distance measures are likely to overestimate the effect of market potential on nominal wages.
ev

32
33 We also found that, according to both approaches, the positive relationship between wages and
34
iew

35 regional market access disappears after 2004. At the same time, individual wages begin to be
36
37 positively correlated with the market potential of Poland’s main trading partners. This clearly
38 demonstrates the importance of international trade in the determination of regional wages in
39
40 Poland.
On

41
42
43 Our results indicate that the theoretical approach by Krugman may not work in the case of
44
ly

45 countries undergoing profound market accessibility changes. As a consequence there might be


46
a need to develop a modified framework that would account for external market potential. Our
47
48 findings also raise an important question concerning the reciprocity of economic influence
49
50 between countries such as Poland and Germany. Hence, future research should focus on issues
51
52 related to the potential impact of international market accessibility improvement in Poland on
53 regional wages in Germany or more broadly on the impact of interregional/intraregional
54
55 accessibility improvement on interregional, intraregional and international trade.
56
57
58
59
60

16
URL: http://mc.manuscriptcentral.com/rsea
Page 17 of 27 Spatial Economic Analysis

1
2
3 References
4
5
6 Alañón, A., Arauzo, J., Myro, R. (2007). Accessibility, agglomeration and location. In J. M.
7 Aruzo, M. Manjón (Eds.), Entrepreneurship, industrial location an economic growth.
8
Cheltenham: Edward Elgar.
9
10 Blanchflower, D., Oswald, A. (1990). The Wage Curve. Scandinavian Journal of Economics
11 92, 215-235.
12
13 Brakman, S., Garretsen, H., Schramm, M. (2004). The spatial distribution of wages and
14 employment: Estimating the Helpman-Hanson model for Germany. Journal of Regional
15 Science 44(3), 437–466.
16
17 Breinlich, H. (2006). The spatial income structure in the European Union. What role for
18 economic geography? Journal of Economic Geography 6(5), 593–617.
Fo

19
20 Cieślik A., Rokicki B., (2016). Individual wages and regional market potential, Economics of
21 Transition, 24(4), 661-682.
rP

22
Combes, P., Mayer, T., Thisse, J. (2008). Economic geography: The integration of regions and
23
24 nations. Princeton: Princeton University Press.
ee

25 Da Mata, D., Deichmann, U., Vernon-Henderson, J., Lall, S., Wang, H. G. (2007).
26 Determinants of city growth in Brazil. Journal of Urban Economics 62(2), 252–272.
27
28 Fally, T., Paillacar, R., Terra, C. (2010). Economic geography and wages in Brazil: Evidence
rR

29 from microdata. Journal of Development Economics 91, 155–168.


30
31 Feenstra, R., Hanson, G. (2000). Aggregation bias in the factor content of trade: Evidence from
ev

32 U.S. manufacturing. American Economic Review 90(2), 150–160.


33
34 Fingleton, B., Palombi, S. (2013). The wage curve reconsidered: is it truly an empirical law of
economics? Région et Développement 38, 49–92.
iew

35
36
Fingleton, B., Palombi, S. (2016). Bootstrap J-Test for Panel Data Models with Spatially
37
38 Dependent Error Components, a Spatial Lag and Additional Endogenous Variables. Spatial
39 Economic Analysis 11(1), 7-26.
40
On

Fujita, M., P. Krugman and A. J. Venables (1999), The Spatial Economy: Cities, Regions and
41
42
International Trade, MIT Press, Cambridge MA.
43 Geurs, K.T., Ritsema van Eck, J.R. (2001). Accessibility Measures: Review and Applications.
44
ly

National Institute of Public Health and the Environment, Bilthoven.


45
46 Geurs, K.T., van Wee, B. (2004). Accessibility evaluation of land-use and transport strategies:
47
48 review and research directions. Journal of Transport Geography 12, 127–140.
49 Hansen, W.G. (1959). How Accessibility Shapes Land Use. Journal of American Institute of
50
Planners 25 (2), 73–76.
51
52 Hanson, G. (2005). Market potential, increasing returns, and geographic concentration. Journal
53 of International Economics 67(1), 1–24.
54
55 Harris, C. (1954). The market as a factor in the localization of industry in the United States.
56 Annals of the Association of American Geographers 64, 315–348.
57
58 Harris, J., Todaro, M. (1970). Migration, unemployment and development: A two-sector
59 analysis. The American Economic Review 60(1), 126-142.
60

17
URL: http://mc.manuscriptcentral.com/rsea
Spatial Economic Analysis Page 18 of 27

1
2
3 Head, K., Mayer, T. (2006). Regional wage and employment responses to market potential in
4
the EU. Regional Science and Urban Economics 36(5), 573–594.
5
6 Hering, L., Poncet, S. (2010). Market access and individual wages: Evidence from China.
7 Review of Economics and Statistics 92(1), 145–159.
8
9 Jiang, H., Xu, W., Zhang, W. (2018). Transportation Accessibility and Location Choice of
10 Japanese-Funded Electronic Information Manufacturing Firms in Shanghai. Sustainability
11 10(2), 1-21.
12
13 Komornicki, T., Rosik, P., Śleszyński, P., Solon, J., Wiśniewski, R., Stępniak, M., Czapiewski,
14 K., Goliszek, S. (2013) Impact of the construction of motorways and expressways on socio-
15 economic and territorial development of Poland. Ministry of Regional Development, Warsaw.
16
17 Krugman, P. (1991a). Increasing returns and economic geography, Journal of Political
18 Economy, 99(3), 483–499.
Fo

19
20
Krugman, P. (1991b). Geography and Trade. Cambridge, MA: MIT Press.
21 Krugman, P. (1992), A Dynamic Spatial Model, NBER Working Papers 4219.
rP

22
23 Lovely, M., Liang, Y., Zhang, H. (2019). Economic geography and inequality in China: Did
24 improved market access widen spatial wage differences? China Economic Review 54, 306-323.
ee

25
26 Matas, A., Raymond, J-L., Roig, J-L. (2015). Wages and Accessibility: The Impact of Transport
27 Infrastructure, Regional Studies 49(7), 1236-1254.
28
rR

Mion, G. (2004). Spatial externalities and empirical analysis: The case of Italy. Journal of
29
30 Urban Economics 56(1), 97–118.
31 Mion, G., Naticchioni, P. (2005). Urbanization externalities, market potential and spatial
ev

32
sorting of skills and firms. CEPR Discussion Paper 5172. Centre for Economic Policy Research,
33
34
London.
iew

35 Östh, J., Reggiani, A., Galiazzo, G. (2014). Novel methods for the estimation of cost–distance
36 decay in potential accessibility models. In: Condeço-Melhorado, A., Reggiani, A., Gutiérrez, J.
37
38
(Eds.), Accessibility and Spatial Interaction. Edward Elgar, Cheltenham.
39 Redding, S. (2013). Economic Geography: A Review of the Theoretical and Empirical
40
On

Literature. In: Bernhofen D., Falvey R., Greenaway D., Kreickemeier U. (eds.) Palgrave
41
Handbook of International Trade. Palgrave Macmillan, London.
42
43 Redding, S., Venables, A. (2004). Economic geography and international inequality. Journal
44 of International Economics 62(1), 53–82.
ly

45
46 Rokicki, B., Hewings, G.J.D. (2019). Regional price deflators in Poland – Evidence from
47 NUTS2 and NUTS3 regions. Spatial Economic Analysis 14 (1), 88–105.
48
49 Rosik, P., Stępniak, M., Komornicki, T. (2015). The decade of the big push to roads in Poland:
50 Impact on improvement in accessibility and territorial cohesion from a policy perspective.
51 Transport Policy 37, 134–146.
52
53 Śleszyński, P. (2014). Expected traffic speed in Poland using Corine land cover, SRTM-3 and
54 detailed population places data. Journal of Maps 11, 245–254.
55
56
57
58
59
60

18
URL: http://mc.manuscriptcentral.com/rsea
Page 19 of 27 Spatial Economic Analysis

1
2
3 Figure 1. Spatial distribution of gross monthly wages in 2015 (Poland = 100)
4
5
6
7
8
9
10
11
12
13
14
15
16
Fo
17
18
19
rP
20
21
22
23
ee

24
25
26
rR

27
28
29
ev

30
31
32
iew

33
34
35
36
37
On

38
39
40
41
ly

42 Source: Authors own analysis.


43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 URL: http://mc.manuscriptcentral.com/rsea
Spatial Economic Analysis Page 20 of 27

1
2
3 Table 1 NLS estimate results from 2000-2015 panel data– individual net monthly wages:
4
5
Krugman approach with different distance measures
6
7 (1) (2) (3)
8 VARIABLES Distance Constant Accessibility
9 between capitals accessibility improvement
10
11 Potential 0.193*** 0.216*** 0.079**
12 (0.035) (0.036) (0.038)
Distance 0.186*** 0.073 -0.003
13
(0.038) (0.053) (0.004)
14
RPP deflators 2.769*** 2.353*** 7.038**
15
(0.727) (0.607) (3.577)
16
Fo
17 Individual characteristics yes yes yes
18 Firm characteristics yes yes yes
19 Regional dummies yes yes yes
rP
20 Time dummies yes yes yes
21
22 Observations 142,112 142,112 142,112
23
ee
Log likelihood -44520.24 -44519.09 -44534.54
24
25 Notes: Robust standard errors in parentheses. *, ** and *** represent significance at 10%, 5% and 1%, respectively.
26
rR

27
28 Table 2. NLS estimate results from 2000-2015 panel data – individual net monthly wages:
29 Krugman approach with accessibility improvement
ev

30
(1) (2) (3) (4) (5)
31
VARIABLES 2000-2015 2002-2015 2004-2015 2006-2015 2008-2015
32
iew

33 Potential 0.079** 0.083 0.085 -0.198 -0.659


34 (0.038) (0.053) (0.105) (0.158) (1.555)
35 Distance -0.003 -0.005 -0.007 -0.006 -0.002
36 (0.004) (0.005) (0.012) (0.007) (0.006)
37 RPP deflators 7.038** 7.408 8.001 -2.578 -0.521
On

38 (3.577) (4.863) (9.734) (2.336) (1.251)


39
40 Individual characteristics yes yes yes yes yes
41 Firm characteristics yes yes yes yes yes
ly

42 Regional dummies yes yes yes yes yes


43 Time dummies yes yes yes yes yes
44
45 Observations 142,112 121,864 103,570 86,114 67,906
46 Log likelihood -44534.54 -38771.37 -33633.33 -28386.99 -22424.07
47
48 Notes: Robust standard errors in parentheses. *, ** and *** represent significance at 10%, 5% and 1%, respectively.
49
50
51
52
53
54
55
56
57
58
59
60 URL: http://mc.manuscriptcentral.com/rsea
Page 21 of 27 Spatial Economic Analysis

1
2
3 Table 3 NLS estimate results from 2000-2015 panel data– individual net monthly wages:
4
5
Krugman approach with accessibility improvement and weighted potential of trade partners
6
7 (1) (2) (3) (4) (5)
8 VARIABLES 2000-2015 2002-2015 2004-2015 2006-2015 2008-2015
9
10 Potential 0.125** 0.161* 0.087 -0.244* -0.527
11 (0.061) (0.083) (0.104) (0.147) (0.496)
Distance -0.003 -0.003 -0.007 -0.007 -0.004
12
(0.002) (0.003) (0.012) (0.006) (0.005)
13
RPP deflators 4.484* 3.842* 7.849 -1.964 -0.604
14
(1.316) (2.088) (9.315) (1.476) (0.697)
15 Trade partners potential -0.044 -0.076 -0.371 1.228** 1.307**
16 (0.043) (0.056) (0.506) (0.586) (0.652)
Fo
17
18 Individual characteristics yes yes yes yes yes
19 Firm characteristics yes yes yes yes yes
rP
20 Regional dummies yes yes yes yes yes
21 Time dummies yes yes yes yes yes
22
23
ee
Observations 142,112 121,864 103,570 86,114 67,906
24 Log likelihood -44533.97 -38770.28 -33633.06 -28384.79 -22422.03
25
26 Notes: Robust standard errors in parentheses. *, ** and *** represent significance at 10%, 5% and 1%, respectively.
rR

27
28 Table 4. OLS estimate results from 2000-2015 panel data – individual net monthly wages:
29 interregional accessibility
ev

30
31
32 (1) (2) (3) (4) (5)
VARIABLES 2000-2015 2002-2015 2004-2015 2006-2015 2008-2015
iew

33
34
Potential accessibility 0.105*** 0.108*** 0.060 -0.076 -0.120*
35
(0.030) (0.036) (0.046) (0.055) (0.068)
36
37 Constant 2.299*** 2.303*** 2.561*** 3.289*** 3.635***
On

38 (0.137) (0.169) (0.210) (0.253) (0.305)


39
40 Individual characteristics yes yes yes yes yes
41 Firm characteristics yes yes yes yes yes
ly

42 Regional dummies yes yes yes yes yes


43 Time dummies yes yes yes yes yes
44
45 Observations 142,112 121,864 103,570 86,114 67,906
46 R-squared 0.48 0.48 0.46 0.42 0.38
47 Adj. R-squared 0.48 0.48 0.46 0.42 0.38
48 Notes: OLS estimation. Robust standard errors in parentheses. *, ** and *** represent significance at 10%, 5% and
49 1%, respectively.
50
51
52
53
54
55
56
57
58
59
60 URL: http://mc.manuscriptcentral.com/rsea
Spatial Economic Analysis Page 22 of 27

1
2
3 Table 5 OLS estimate results from 2000-2015 panel data – individual net monthly wages:
4
5
interregional accessibility potential of with trade partners (weighted)
6
7 (1) (2) (3) (4) (5)
8 VARIABLES 2000-2015 2002-2015 2004-2015 2006-2015 2008-2015
9
10 Potential accessibility 0.104*** 0.096*** 0.060 -0.095* -0.146**
11 (0.031) (0.037) (0.046) (0.056) (0.069)
Trade partners potential -0.023 -0.488 -0.286 1.169** 1.127*
12
(0.221) (0.307) (0.489) (0.564) (0.602)
13
14
Individual characteristics yes yes yes yes yes
15 Firm characteristics yes yes yes yes yes
16 Regional dummies yes yes yes yes yes
Fo
17 Time dummies yes yes yes yes yes
18
19 Constant 2.525 7.007** 5.323 -7.744 -6.845
rP
20 (2.172) (2.967) (4.725) (5.331) (5.613)
21
22 Observations 142,112 121,864 103,570 86,114 67,906
23
ee
R-squared 0.48 0.48 0.46 0.42 0.38
24 Adj. R-squared 0.48 0.48 0.46 0.42 0.38
25 Notes: OLS estimation. Robust standard errors in parentheses. *, ** and *** represent significance at 10%, 5% and
26 1%, respectively.
rR

27
28 Table 6. OLS estimate results from 2000-2015 panel data – individual net monthly wages:
29
interregional accessibility with potential of trade partners(weighted) and accessibility spillover
ev

30
31
32 (1) (2) (3) (4) (5)
iew

33 VARIABLES 2000-2015 2002-2015 2004-2015 2006-2015 2008-2015


34
35 Interregional accessibility -0.064* -0.033 -0.004 -0.091 -0.155**
(0.035) (0.039) (0.046) (0.056) (0.070)
36
Accessibility spillover 0.264*** 0.305*** 0.385*** 0.170 -0.265*
37
(0.053) (0.062) (0.082) (0.113) (0.161)
On

38
Trade partners potential 1.421*** 1.322*** 8.709*** 2.149*** 0.905***
39 (0.050) (0.061) (0.557) (0.154) (0.127)
40
41 Constant -13.289*** -12.519*** -93.463*** -20.548*** -4.789***
ly

42 (0.378) (0.439) (5.690) (1.128) (0.583)


43
44 Individual characteristics yes yes yes yes yes
45 Firm characteristics yes yes yes yes yes
46 Regional dummies yes yes yes yes yes
47 Time dummies yes yes yes yes yes
48
49 Observations 142,112 121,864 103,570 86,114 67,906
50 R-squared 0.48 0.47 0.45 0.42 0.37
51 Adj. R-squared 0.48 0.47 0.45 0.42 0.37
52 Notes: OLS estimation. Robust standard errors in parentheses. *, ** and *** represent significance at 10%, 5% and
53 1%, respectively.
54
55
56
57
58
59
60 URL: http://mc.manuscriptcentral.com/rsea
Page 23 of 27 Spatial Economic Analysis

1
2
3 Table 7 Results of estimates of instrumental variables from 2000-2015 panel data – individual
4
5
net monthly wages: interregional accessibility with trade partners’ potential (weighted) – inverse
6 distance
7
8 (1) (2) (3) (4) (5)
9 VARIABLES 2000-2015 2002-2015 2004-2015 2006-2015 2008-2015
10
11 Interregional accessibility 0.120*** 0.130*** 0.056 -0.107 -0.290***
12 (0.034) (0.044) (0.061) (0.076) (0.106)
13 Trade partners potential 1.496*** 1.452*** 10.615*** 2.363*** 0.818***
14 (0.041) (0.049) (0.400) (0.096) (0.075)
15
16 Constant -13.743*** -13.284*** -112.772*** -22.015*** -4.375***
Fo
(0.330) (0.370) (4.109) (0.749) (0.405)
17
18
Individual characteristics yes yes yes yes yes
19
Firm characteristics yes yes yes yes yes
rP
20 Regional dummies yes yes yes yes yes
21 Time dummies yes yes yes yes yes
22
23
ee

Observations 142,112 121,864 103,570 86,114 67,906


24 R-squared 0.48 0.48 0.46 0.42 0.38
25 Adj. R-squared 0.48 0.48 0.46 0.42 0.38
26
rR

Notes: FE-2SLS estimation. Robust standard errors in parentheses. *, ** and *** represent significance at 10%, 5%
27 and 1%, respectively.
28
29
ev

30
31
32
iew

33
34
35
36
37
On

38
39
40
41
ly

42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 URL: http://mc.manuscriptcentral.com/rsea
Spatial Economic Analysis Page 24 of 27

1
2
3 Data Appendix
4
5
6 The data set used in the present study is based on the fourth quarter of the Polish Labor Force
7 Survey (LFS) for the period 2000-20151. The survey does not gather information on wages of
8
9 self-employed or paid family workers. After excluding the unemployed, inactive and missing
10
11 observations, we are left with over 142,112 observations.
12
13
14 The data from the Polish Labor Force Survey provides information on the region (at the NUTS2
15 level) where the employee is located and the size of the city where he or she lives. Furthermore,
16
17 the Survey gives data on personal characteristics, employment spell and employer. In this study
18
Fo

19 we use individual net hourly wage deflated by the national consumer price index. We also apply
20
21
the unemployment rate calculated on the basis of the LFS data, taken from the Polish Central
rP

22 Statistical Office for the population aged 15+.


23
24
ee

25 The control variables used in the regressions are the following:


26
27
28  Age of the individual
rR

29
30  Gender. Male = 0 and female = 1.
31

ev

32 Marital status. This variable includes 4 categories: Married, Single, Widow, Divorced.
33
34
 Education. This variable includes 7 different categories: higher (including second stage of
iew

35 tertiary education) = 7, post-secondary = 6, secondary vocational = 5, general secondary =


36
37 4, vocational = 3, primary and lower secondary = 2, incomplete primary and no education
38
39 = 1.
40
On

 Region. We distinguish among 16 Polish NUTS2 regions.


41
42  City size. This variable includes 8 categories: city with population over 100 thousand = 7,
43
44 50 – 99.9 thousand = 6, 20 – 49.9 thousand = 5, 10 – 19.9 thousand = 4, 5 – 9.9 thousand
ly

45
46 = 3, 2 – 4.9 thousand = 2, less than 2 thousand = 1, village = 0.
47
48
 The individual's years of tenure at the firm.
49  Type of work. Permanent contract = 0, temporary contract = 1.
50
51  Occupation. This variable includes 9 categories: representatives of the public authorities,
52
53 senior officials and managers, specialists, technicians and other medium staff, office staff,
54
55 personal services staff and dealers, farmers, gardeners, foresters and fishermen, industrial
56
57
58
59
60 1 The survey is being conducted quarterly, so there are 4 data sets each year.

1
URL: http://mc.manuscriptcentral.com/rsea
Page 25 of 27 Spatial Economic Analysis

1
2
3 workers and craftsmen, operators and assemblers of machinery and equipment, employed
4
5 in elementary occupations.
6
7  Industry classification. This variable defines the specific industry to which the employing
8
9
establishment belongs, in accordance to the NACE rev. 1.1 classification. These include
10 15 categories such as: primary sector, mining and quarrying, manufacturing, electricity,
11
12 gas and water supply, construction, trade and repair, hotels and restaurants, transport,
13
14 storage and communication, financial intermediation, real estate, renting and business
15
activities, public administration, education, health care and social work, other community,
16
17 social and personal service activities, households.
18
Fo

19  Firm size. This variable includes 5 categories: up to 10 employees = 1, 11-19 employees =


20
21 2, 20-49 employees = 3, 50-100 employees = 4, over 100 employees = 5.
rP

22
 Sector. Public = 0, private = 1.
23
24
ee

25 Summary statistics
26
27
28
rR

Variable Obs Mean Std. Dev. Min Max


29
30
31 Region 142,112 16.913 9.191 2 32
ev

32 Year 142,112 2007.386 4.754 2000 2015


33 Gender 142,112 0.479 0.500 0 1
34
Age 142,112 39.844 11.163 15 74
iew

35
36 Age squared 142,112 1712.175 909.149 225 5476
37 Marital status 142,112 0.398 0.746 0 3
38 Net monthly wage 142,112 1472.193 884.668 100 22000
39
40 Working hours 142,112 39.776 7.400 1 60
On

41 Section 142,112 10.085 3.967 2 18


42 Firm size 142,112 3.373 1.512 1 5
43 Tenure 142,112 9.372 9.608 0 53
44
ly

45 Tenure squared 142,112 180.134 304.436 0 2809


46 City size 142,112 3.460 2.945 0 7
47 Occupation 142,112 3.698 2.522 0 8
48 Sector 142,112 0.617 0.486 0 1
49
50 Type of work 142,112 0.241 0.428 0 1
51 Education 142,112 4.509 1.658 1 7
52
53
54
55
56
57
58
59
60

2
URL: http://mc.manuscriptcentral.com/rsea
Spatial Economic Analysis Page 26 of 27

1
2
3 Appendix 1. External market potential
4
5
6 Table A1. Estimation results panel data 2000-2015 – individual net monthly wages:
7 interregional accessibility with trade partners potential (not weighted) – inverse distance to
8 Brussels
9
10
(1) (2) (3) (4) (5)
11
VARIABLES 2000-2015 2002-2015 2004-2015 2006-2015 2008-2015
12
13
Interregional accessibility 0.105*** 0.108*** 0.060 -0.076 -0.120*
14
(0.030) (0.036) (0.046) (0.055) (0.068)
15
Trade partners potential 1.512*** 1.473*** 10.590*** 2.326*** 0.705***
16 (0.037) (0.042) (0.316) (0.074) (0.051)
17
18
Fo

Constant -13.845*** -13.422*** -112.517*** -21.756*** -3.899***


19 (0.311) (0.333) (3.265) (0.609) (0.337)
20
21 Individual characteristics yes yes yes yes yes
rP

22 Firm characteristics yes yes yes yes yes


23 Regional dummies yes yes yes yes yes
24 Time dummies yes yes yes yes yes
ee

25
26 Observations 142,112 121,864 103,570 86,114 67,906
27 R-squared 0.48 0.48 0.46 0.42 0.38
28 Adj. R-squared 0.48 0.48 0.46 0.42 0.38
rR

29 Notes: OLS estimation. Robust standard errors in parentheses. *, ** and *** represent significance at 10%, 5%
30 and 1%, respectively.
31
ev

32 Table A2. Instrumental variables estimation results panel data 2000-2015 – individual net
33 monthly wages: interregional accessibility with trade partners potential (not weighted) –
34
inverse distance to Brussels
iew

35
36
37 (1) (2) (3) (4) (5)
38 VARIABLES 2000-2015 2002-2015 2004-2015 2006-2015 2008-2015
39
Interregional accessibility 0.120*** 0.130*** 0.056 -0.107 -0.290***
40
On

(0.034) (0.044) (0.061) (0.076) (0.106)


41
Trade partners potential 1.496*** 1.452*** 10.615*** 2.363*** 0.818***
42
(0.041) (0.049) (0.400) (0.096) (0.075)
43
44
ly

Constant -13.743*** -13.284*** -112.772*** -22.015*** -4.375***


45 (0.330) (0.370) (4.109) (0.749) (0.405)
46
47 Individual characteristics yes yes yes yes yes
48 Firm characteristics yes yes yes yes yes
49 Regional dummies yes yes yes yes yes
50 Time dummies yes yes yes yes yes
51
52 Observations 142,112 121,864 103,570 86,114 67,906
53 R-squared 0.48 0.48 0.46 0.42 0.38
54 Adj. R-squared 0.48 0.48 0.46 0.42 0.38
55 Notes: FE-2SLS estimation. Robust standard errors in parentheses. *, ** and *** represent significance at 10%,
56 5% and 1%, respectively.
57
58
59
60

3
URL: http://mc.manuscriptcentral.com/rsea
Page 27 of 27 Spatial Economic Analysis

1
2
3 Appendix 2. Special case – Germany as the main Poland’s trading partner.
4
5
6
7
Table A1. Estimation results panel data 2000-2015 – individual net monthly wages:
8 interregional accessibility with German potential – inverse distance to Hannover
9
10 (1) (2) (3) (4) (5)
11 VARIABLES 2000-2015 2002-2015 2004-2015 2006-2015 2008-2015
12
13 Interregional accessibility 0.105*** 0.108*** 0.060 -0.076 -0.120*
14 (0.030) (0.036) (0.046) (0.055) (0.068)
15 German potential 1.551*** 1.400*** 4.755*** 1.735*** 0.600***
16 (0.038) (0.040) (0.142) (0.055) (0.044)
17
18 Constant -13.512*** -11.935*** -46.588*** -14.459*** -2.465***
Fo

19 (0.303) (0.293) (1.301) (0.387) (0.251)


20
Individual characteristics yes yes yes yes yes
21
rP

Firm characteristics yes yes yes yes yes


22
Regional dummies yes yes yes yes yes
23 Time dummies yes yes yes yes yes
24
ee

25 Observations 142,112 121,864 103,570 86,114 67,906


26 R-squared 0.48 0.48 0.46 0.42 0.38
27 Adj. R-squared 0.48 0.48 0.46 0.42 0.38
28
rR

Notes: OLS estimation. Robust standard errors in parentheses. *, ** and *** represent significance at 10%, 5%
29 and 1%, respectively.
30
31
ev

32
33 Table A2. Instrumental variables estimation results panel data 2000-2015 – individual net
34 monthly wages: interregional accessibility with German potential – inverse distance to
iew

35 Hannover
36
37 (1) (2) (3) (4) (5)
38 VARIABLES 2000-2015 2002-2015 2004-2015 2006-2015 2008-2015
39
40 Interregional accessibility 0.120*** 0.130*** 0.056 -0.107 -0.290***
On

41 (0.034) (0.044) (0.061) (0.076) (0.106)


42 German potential 1.535*** 1.379*** 4.767*** 1.762*** 0.696***
43 (0.042) (0.046) (0.180) (0.072) (0.064)
44
ly

45 Constant -13.414*** -11.818*** -46.687*** -14.604*** -2.711***


46 (0.322) (0.323) (1.621) (0.458) (0.276)
47
48 Individual characteristics yes yes yes yes yes
49 Firm characteristics yes yes yes yes yes
50 Regional dummies yes yes yes yes yes
51 Time dummies yes yes yes yes yes
52
Observations 142,112 121,864 103,570 86,114 67,906
53
R-squared 0.48 0.48 0.46 0.42 0.38
54
Adj. R-squared 0.48 0.48 0.46 0.42 0.38
55
Notes: FE-2SLS estimation. Robust standard errors in parentheses. *, ** and *** represent significance at 10%,
56
5% and 1%, respectively.
57
58
59
60

4
URL: http://mc.manuscriptcentral.com/rsea

You might also like