Geographically Weighted Regression In The Analysis Of . - MDPI

ISPRS Int. J. Geo-Inf. 2018, 7, 17 3 of 16 ISPRS J. Geo-Inf. 2018, 7, 17 level Int. 2 (2478 units), therefore 3 of 16 these units were not analysed in the study) in Poland (Projected Coordinate System: ETRS89 Poland CS92). This study study consists consists of in in Poland. It This of six six parts. parts. Section Section22introduces introducesthe theissue issueofofunemployment unemployment Poland. presents a preliminary statistical data analysis using GIS and spatial statistics tools. Section It presents a preliminary statistical data analysis using GIS and spatial statistics tools. Section 33 discusses economic economic theories theories on on unemployment unemployment determinants determinants and and presents presents the the final final data data set set used used in in discusses the study. The fourth section describes the method applied in the analysis of unemployment in the study. Section 4 describes the method applied in the analysis of unemployment in Poland—GWR. Poland—GWR. It also contains a short review of research based on GWR. Section 5 discusses the It also contains a short review of research based on GWR. Section 5 discusses the results of the analyses results of thedeterminants analyses of the local determinants of unemployment Poland. Based on the obtained of the local of unemployment in Poland. Based on in the obtained estimation results, estimation results, the differences between the global (ordinary least squares) and local (GWR) the differences between the global (ordinary least squares) and local (GWR) models are identified. models identified. Theafinal section provides summary and general conclusions. The finalare section provides summary and generalaconclusions. 2. Unemployment Unemployment Rate Rate in in Poland Poland 2. The unemployment unemployment rate rate in in Poland Poland has has varied varied considerably considerably since since 1990. 1990. There There have have been been periods periods The when the rate has increased (1991–1993, 1999–2002, 2009–2012) and fallen (1994–1996, 2004–2008, when the rate has increased (1991–1993, 1999–2002, 2009–2012) and fallen (1994–1996, 2004–2008, 2014). 2014). Poland’s unemployment rate peaked 2002 and and 2003,then andfell then to aoflow of in 9.5% in Poland’s unemployment rate peaked at 20%atin20% 2002inand 2003, to fell a low 9.5% 2008. 2008. In it2015, was(see 9.7%Figure (see Figure 1). Changes the unemployment level closely were closely connected In 2015, wasit9.7% 1). Changes to theto unemployment level were connected with with the country’s economic and political situation. It was strongly affected by, for example, the country’s economic and political situation. It was strongly affected by, for example, the expirythe of expiry of obligations in the privatisation agreements of the mid-1990s, which required obligations set forth in set the forth privatisation agreements of the mid-1990s, which required enterprises to enterprises to maintainatemployment at a specified levelFurther (1998–2003). Further influences periods maintain employment a specified level (1998–2003). influences were periods were of economic of economic growth (1994–1997, 2004–2008), economic crisis (after 2008), and mass migration for growth (1994–1997, 2004–2008), economic crisis (after 2008), and mass migration for economic reasons economic reasons connected with Poland’s accession to the European connected with Poland’s accession to the European Union (after 2004). Union (after 2004). 22 20% 20% 20 18 16 14 12 2015 2013 2012 2011 9.7% 2010 2009 2007 2006 2005 2004 2003 2002 2001 2000 1999 1998 1997 1996 1995 1994 1993 1992 1991 2008 9.5% 8 2014 10 Figure Figure 1. 1. Unemployment Unemployment rate rate in in Poland Poland (1990–2015). (1990–2015). Source: Source: Own Own elaboration. elaboration. The level level of of unemployment unemployment in in Poland Poland has has shown shown considerable considerable spatial spatial (local (local and and regional) regional) The diversification. In In 2015, 2015, the the difference difference between between districts districts (LAU1), (LAU1), characterised characterised by by the the highest highest (the (the diversification. Szydlowiecki district: 30.8%) and lowest (the town of Poznan with district rights: 2.4%) Szydlowiecki district: 30.8%) and lowest (the town of Poznan with district rights: 2.4%) unemployment unemployment as large as 28.4 Atdifference the voivodeship level, the levels, was as largelevels, as 28.4was percentage points. At thepercentage voivodeshippoints. level, the was 10.1 percentage difference was 10.1 percentage points (between Warminsko-Mazurskie voivodeship with 16.2% and points (between Warminsko-Mazurskie voivodeship with 16.2% and Wielkopolskie with 6.1%). Wielkopolskie with 6.1%). The maps presented in Figure 2 show that northern Poland has the highest unemployment rate. The maps presented in Figure 2 showinthat northern PolandWarsaw, has the highest unemployment rate. In contrast, the lowest values are observed large cities: Poznan, Katowice, Krakow, Wroclaw In contrast, the lowest values are observed in large cities: Poznan, Warsaw, Katowice, Krakow, and the Tricities (Gdansk, Gdynia and Sopot). Also of note, the districts located in the Mazowieckie Wroclaw andwere the among Tricitiesthose (Gdansk, Gdynia and Sopot). Also of and note, the districts located in the voivodeship characterised by both the highest lowest unemployment rates, Mazowieckie among those characterised byamong both the highest lowest indicating the voivodeship considerable were diversity of economic development spatial unitsand located in unemployment rates, indicating the considerable diversity of economic development among spatial that voivodeship. units located in that voivodeship.

ISPRS Int. J. Geo-Inf. 2018, 7, 17 4 of 16 ISPRS Int. J. Geo-Inf. 2018, 7, 17 ISPRS Int. J. Geo-Inf. 2018, 7, 17 4 of 16 4 of 16 Figure 2. 2. Unemployment rates in bydistrict district (LAU and voivodeship (NUTS in Figure rates in Poland Poland by (LAU 1)1) and (NUTS 2) in2) 2015. Source: Figure 2.Unemployment Unemployment rates in Poland by district (LAU 1)voivodeship and voivodeship (NUTS 2) 2015. in 2015. Source: Own elaboration in ArcMap 10.2. Own elaboration in ArcMap 10.2. Source: Own elaboration in ArcMap 10.2. Figure thetheresults of ofgrouping districts to tounemployment rates. Figure33 shows 3shows shows results grouping districtsaccording unemployment rates. Figure the results of grouping districts according toaccording unemployment rates. Unemployment Unemployment rates below the national mean (9.7%) were observed in 139 districts, mainly in towns Unemployment rates below the national mean (9.7%) were observed in 139 districts, mainly in towns rates below the national mean (9.7%) were observed in 139 districts, mainly in towns with district with district rights. Values fluctuated around the natural rate of of unemployment (below 6%) in in 38 38 with district Values fluctuated around natural rate unemployment 6%) rights. Valuesrights. fluctuated around the natural ratethe of unemployment (below 6%) in 38(below districts (which districts (which represents 10% of all analysed units). The biggest group consists of districts where districts (which 10% of allThe analysed biggest group consists ofvariable districtsranged where represents 10% ofrepresents all analysed units). biggestunits). group The consists of districts where the thethe variable ranged from 9.7% to 15%, representing 139 spatial units. Unemployment rates above from 9.7% 139 to 15%, representing 139 spatial rates units.above Unemployment rates above fromvariable 9.7% to ranged 15%, representing spatial units. Unemployment 20% are observed in as 20% areare observed in as many as as 36 36 districts. In oneone of these, the rate exceeded 30%. 20% in In as many districts. of these, many as observed 36 districts. one of these, the rate In exceeded 30%. the rate exceeded 30%. 139139 140140 120120 101101 100100 80 80 60 60 40 40 66 66 35 35 38 38 20 20 0 0 1 1 under 6%6% under 6.1-9.6% 6.1-9.6% 9.7-15% 9.7-15% 15.5-20% 15.5-20% 20.1-30% than 30% 20.1-30% more more than 30% Figure 3. Districts grouped according to to unemployment rate in in 2015. Source: Own elaboration. Figure 3. grouped according rate Source: Own elaboration. Figure 3. Districts Districts grouped according to unemployment unemployment rate in2015. 2015. Source: Own elaboration. The Moran’s I statistic forfor the unemployment rate in in 2015 was 0.47 (for(for thethe spatial weights The Moran’s statistic unemployment was 0.47 spatial The Moran’s I Istatistic for thethe unemployment raterate in 20152015 was 0.47 (for the spatial weightsweights matrix matrix in the queen configuration, the result was statistically significant). This means that matrix in the queen configuration, the result was statistically significant). This means in the queen configuration, the result was statistically significant). This means that unemploymentthat in unemployment in Poland was characterised byby a relatively high and positive spatial autocorrelation. unemployment in Poland was characterised apositive relatively high and positive spatial autocorrelation. Poland was characterised by a relatively high and spatial autocorrelation. Moreover, there were Moreover, there were spatial relationships among thethe districts that affected thethe unemployment rates. Moreover, there were spatial relationships among districts that affected rates. spatial relationships among the districts that affected the unemployment rates. unemployment Therefore, clusters of Therefore, clusters of districts occurred in the geographic space, characterised by similar unemployment Therefore, clustersinofthe districts occurred in the geographic space, characterised by similar unemployment districts occurred geographic space, characterised by similar unemployment rates [27]. rates [27]. rates [27].local Moran’s statistic answers the question as to where exactly in the analysed area this The The local Moran’s statistic answers the question as to to where exactly in in thethe analysed area thisthis The local Moran’s statistic answers question where exactly analysed area phenomenon arose. In Figure 4, the areas the coloured greyasindicate clusters of districts characterised by phenomenon arose. In In Figure 4, the areas coloured grey indicate clusters of districts characterised by phenomenon Figure 4,rates. the areas coloured grey indicate of districts characterised similarly higharose. unemployment It is clearly visible that theyclusters are located in the northern part by of similarly high unemployment rates. It is visible that they areare located in in thethe northern part of of similarly high unemployment rates. It clearly is clearly visible that they located northern part the country (Zachodniopomorskie, Warminsko-Mazurskie and Kujawsko-Pomorskie voivodeships). thethe country (Zachodniopomorskie, Warminsko-Mazurskie and Kujawsko-Pomorskie voivodeships). country Warminsko-Mazurskie and Kujawsko-Pomorskie voivodeships). In turn, the (Zachodniopomorskie, black areas denote clusters of spatial units with similarly low unemployment rates. In In turn, the black areas denote clusters of spatial units with similarly low unemployment rates. They turn, blackinareas denote clusters of spatial units with similarly low(Mazowieckie unemployment rates. They They arethe located cities: Poznan (Wielkopolskie voivodeship), Warsaw voivodeship) areare located in in cities: Poznan (Wielkopolskie voivodeship), Warsaw (Mazowieckie voivodeship) and cities: Poznan (Wielkopolskie voivodeship), Warsaw (Mazowieckie voivodeship) and and located Katowice (Slaskie voivodeship). Statistically non-significant results were obtained for the other Katowice (Slaskie voivodeship). Statistically non-significant results were obtained forfor thethe other Katowice (Slaskie voivodeship). Statistically non-significant results were obtained other districts (i.e., there are no spatial relationships). districts (i.e., there areare nono spatial relationships). districts (i.e., there spatial relationships).

ISPRS Int. J. Geo-Inf. 2018, 7, 17 ISPRS Int. J. Geo-Inf. 2018, 7, 17 5 of 16 5 of 16 Figure 4. Local Moran’s statistic for unemployment rates in Poland in 2015. Source: Own elaboration Figure 4. Local Moran’s statistic for unemployment rates in Poland in 2015. Source: Own elaboration in GeoDa. in GeoDa. Some important conclusions can be drawn from the above analysis. It shows that there were some relationships among affecting the above values analysis. of the unemployment rate in Poland. Some spatial important conclusions candistricts be drawn from the It shows that there were some Thus, information about the relationships among the analysed spatial units ought to be considered spatial relationships among districts affecting the values of the unemployment rate in Poland. Thus, in an econometric model describing unemployment figures in the country, for example, in the form information about the relationships among the analysed spatial units ought to be considered in an of a spatial weights matrix. econometric model describing unemployment figures in the country, for example, in the form of a spatial matrix. 3. weights Determinants of Regional Unemployment Unemployment is a Unemployment multidimensional phenomenon determined by complex factors and 3. Determinants of Regional mechanisms. There are numerous theories offered in the literature that specify the variables which Unemployment multidimensional phenomenon determined byPhillips complex cause an increase is or adecrease in the unemployment rate [1]. According to [28],factors higher and mechanisms. There are numerous theories offered in the literature that specify the variables which unemployment rates are accompanied by a slower rise in nominal wages, while a fall in with an increase in nominal wages. The unemployment rate[28], is also causeunemployment an increase orcoincides decrease in the unemployment rate [1]. According to Phillips higher strongly related Domestic by Product (on rise the in regional level Grosswhile Regional Product). When unemployment rates to areGross accompanied a slower nominal wages, a fall in unemployment employment rises, and thusThe economic growth occurs with simultaneous in to coincides with anincreases, increasethe inGDP nominal wages. unemployment rate is aalso strongly fall related the unemployment rate. Okun’s law states that every 2% drop in real GDP, as compared with Gross Domestic Product (on the regional level Gross Regional Product). When employment increases, potential GDP, results in a rise in the unemployment rate by 1 percentage point [29]. The number of the GDP rises, and thus economic growth occurs with a simultaneous fall in the unemployment created jobs is largely dependent on the volume and type of investments. New rate. Okun’s law states that every 2% drop in real GDP, as compared with potential GDP, results in development-oriented investments contribute to an increased demand for labour. In contrast, a risecurrent in the investments unemployment by 1 percentage point [29]. The number created jobs is largely aimedrate at property replacement enable existing jobs to beof maintained. It should dependent on the volume and type of investments. New development-oriented investments contribute be emphasised that not all investments contribute to creating or maintaining jobs as they increase the to an productivity increased demand for labour. contrast,greater current investments aimed at property replacement of the workforce [30].InHowever, workforce productivity enables enterprises to operate more effectively, hence, making them more competitive,that and not this all improves the employment enable existing jobs to be maintained. It should be emphasised investments contribute to situation in the longjobs run.as If they productivity significantly, it will increase[30]. the growth rate greater of creating or maintaining increaseincreases the productivity of the workforce However, GDP at a higher rate than productivity, which forces employers at that point to hire more workers to workforce productivity enables enterprises to operate more effectively, hence, making them more accommodate the expected demand. Wages will rise but if labour productivity increases at a rate competitive, and this improves the employment situation in the long run. If productivity increases faster than the increase in wages, then the rates of inflation and unemployment will decline [1]. significantly, it will increase the growth rate of GDP at a higher rate than productivity, which forces Foundations for regional unemployment analysis are set up by the neoclassical theory, which employers at that hire run more to accommodate the due expected demand. Wages will rise suggests that point in thetolong allworkers disparities should disappear to labour or capital flows. but if Therefore, labour productivity increases at a rate faster than the increase in wages, then the rates of inflation another important factor influencing the level of unemployment is migration. The most and unemployment will decline frequently mentioned motives[1]. behind migration include high unemployment rates, low wages and Foundations for regional analysis up by the neoclassical high costs of living. Accordingunemployment to the world systems theory are (it is set a concept of social development theory, in which the erstwhile analysed units, that is the state, economy, society, have been replaced by which suggests that in the long run all disparities should disappear due to labour or capital flows. historical systems; the world is considered as a the spatio-temporal whole), migration results from anmost Therefore, another important factor influencing level of unemployment is migration. The economic imbalance between the core (i.e., developed areas-countries, regions) and peripheries (i.e., frequently mentioned motives behind migration include high unemployment rates, low wages and developing areas that constitute workforce reserves for the core). Since Poland’s accession to the high costs of living. According to the world systems theory (it is a concept of social development in European Union, over one million people have left the country, which resulted in a steadily which the erstwhile analysed units, that is the state, economy, society, have been replaced by historical systems; the world is considered as a spatio-temporal whole), migration results from an economic imbalance between the core (i.e., developed areas-countries, regions) and peripheries (i.e., developing

ISPRS Int. J. Geo-Inf. 2018, 7, 17 6 of 16 areas that constitute workforce reserves for the core). Since Poland’s accession to the European Union, over one million people have left the country, which resulted in a steadily declining unemployment rate from 2004 to 2008 (Figure 1). However, the unemployment rate depends not only on external migration but also, in large measure, on internal-interregional and intraregional ones. The literature suggests that propensity to migrate depends on the level of education of the employees. The highly skilled workers migrate promptly in response to a decline in regional labor demand, while the low-skilled drop out of the labor force or stay unemployed [31]. Regional studies indicate that there are some specific factors affecting regional unemployment differentials. First, conditions in a specific (regional, local) labour market that can be described by a population’s structure (e.g., the number of people of working age, number of men and women), economic activity, level of education and qualifications of local labour force, as well as the number of registered economic entities and offered jobs. Second, regional specialisation (agricultural, industrial and service) and market potential [32]. Among the factor that affect unemployment are also those of a social nature, such as social security policy (e.g., unemployment insurance benefits, family allowances). The correlation between these two factors is positive because the extensive benefits system deters job search and reduces the likelihood of migration [32]. Regrettably, not all variables significant to an analysis of unemployment were available from Polish public statistics for a LAU 1 spatial cross-section. These included, for example, inflation, GDP, level of education or regional specialisation. Because the GDP variable (describing the level of economic development) was important for this analysis, it was replaced with a local development measure which is commonly used in local analysis: districts’ budgetary income per capita. Districts’ budgetary income per capita is not as good a measure of economic development as the GDP that presents total value of goods and services produced in a country or a specific region (it aims to best capture the true monetary value of economy). Nevertheless, the amount of districts’ budgetary income depends not only on general subsidies from the state budget but also on their own revenues from local governments which include e.g., incomes from

The main aim of this article is an application of geographically weighted regression (GWR) in the analysis of unemployment in Poland on the local level (LAU 1) in 2015. GWR enables to identify the variability of regression coefficients within the geographic space. Therefore, the analysis results will allow us to answer all above study question.