The impact of territorial capital on Cohesion Policy in rural Polish areas

ABSTRACT This study evaluates the outcomes of the European Union’s Cohesion Policy on rural areas in Poland between 2007 and 2018. Stratified propensity score matching was used to examine whether policy impact depends on the type of local territorial capital involved. Inputs and outcomes between individual municipalities with similar territorial capital characteristics and conditions of socio-economic development were modelled and compared. The results show that more advanced development of non-agricultural functions and also a reduced agricultural function led to greater Cohesion Policy effectiveness. Thus, convergence, or the evening out of differences in socio-economic development, does not take place at the local (intraregional) level.


INTRODUCTION
Since joining the European Union (EU) in 2004, Poland has been the biggest beneficiary of Cohesion Policy (CP) funds, with €67 billion invested in the 2007-13 programme period (European Commission, 2019), and over €90 billion contracted for 2014-20 (European Commission, 2020). These high sums reflect Poland's high population and low economic development compared with the EU average. Although 2004-18 was a period of continuous and dynamic development (in terms of growth of gross domestic product (GDP) per capita), economic and social disparities have not actually decreased within the country, and may even have deepened. Traditionally, rich and urbanized regions have developed faster than those that are rural and relatively less developed, despite spatially aligned investment (Heffner & Gibas, 2015;Gagliardi & Percoco, 2017). The territorial resources of these areas vary significantly, and nowadays an emerging selective pattern of growth can be seen in each unit of territory (Camagni & Capello, 2013). In this study, following Capello and Dentinho (2012), we define territorial capital as the set of resources that determine a territory's competitive potential.
CP includes the EU's main instruments for supporting regions' and member states' economic and social development (Bachtler & Wren, 2006). The impact of CP funding on rural development has not been comprehensively researched, despite its importance for the evaluation and planning of successive programme periods. Individual evaluation studies have focused on fragmented issues, looking at the impact of separate programmes on individual components of territorial capital, such as entrepreneurship, human capital, infrastructure or quality of life. Yet often these studies do not distinguish between rural and urban areas (Ledzion et al., 2016;Penszko, 2017;Rosik et al., 2015;Wolański et al., 2018). This has created a need for much more complex research of the overall CP impact on rural development in Poland from a complex social and economic perspective, taking account of the spatial diversity of municipalities' 1 territorial capital, but excluding towns and cities. The idea that the effect intensity of CP depends on the type of local territorial capital was tested. Econometric modelling was used to examine the relationships between the diversity of CP intervention in municipality types with similar developmental characteristics (i.e., territorial capital characteristics), and the conditions of socio-economic development. The following hypotheses were proposed: (1) CP has a greater impact on economic than social conditions in rural development; (2) the local economy structure is the main factor determining the effect of CP interventions; and (3) the impact of CP is greatest in the suburban municipalities of Poland's biggest cities.
Stratified propensity score matching (SPSM) analysis enabled significant conclusions regarding the CP's effect on the development of 2174 rural and urban-rural municipalities for the years 2007-18, grouped according to territorial capital characteristics. The study data were obtained from the database of the Ministry of Development Funds and Regional Policy, which is responsible for the management and evaluation of Polish cohesion funds from the EU budget, and from the Rural Development Monitoring study (MROW), 2 which obtains statistical information that is not published by Statistics Poland (GUS), on all Poland's municipalities and drawn from many public institutions.
The study contributes to the discussion on how effective EU CP is in rural development. It focuses on assessing the long-term impact of CP investment in rural areas (covering all NUTS-5/local-level units in a specific country), which is a pioneering approach in studies of this type. Furthermore, evaluative research conducted at the local level (below NUTS-3) is rare. As a result, the study helps to clarify the importance of European Cohesion Funds not only for Poland, but also other countries in Central-Eastern Europe that joined the EU together with Poland in 2004.
The remainder of the paper is structured as follows. Section 2 presents an overview of the literature, drawing attention to the reasons behind the varied results of existing studies on the effects of CP, and the factors explaining CP effectiveness. Section 3 presents indicators selected from the extensive empirical material and variety of sources of available data, together with their basic descriptive statistics. Section 4 outlines the innovative SPSM modelling method used in this study. The empirical evidence related to the investment effects of EU funding in different types of municipalities is discussed in Section 5. Section 6 discusses the results of the study together with the findings of other research, and presents recommendations for territorial development policies and further research.
Existing studies indicate that the varied results of research on the effectiveness of CP measures may have different causes. First could be the great complexity of the interventions involved. Especially, the extensive range of support measures affects different policy areas, from transport infrastructure through to social infrastructure and lifelong learning programmes, and each come with various time lags (Rodríguez-Pose & Fratesi, 2004). Second, the explanation could lie in the time range of the analysis. Importantly, studies focusing on later programme periods tend to indicate a greater and more noticeable CP effectiveness, because of the learning that has taken place over time and the continuous improvement of policymakers (Pinho et al., 2015;Rodríguez-Pose & Novak, 2013). Finally, another likely explanation may be the excessively large or general unit of analysis that is used. So far, the territorial scope of existent studies has almost exclusively encompassed countries and regions (NUTS-1, NUTS-2) (Bradley, 2006;Mohl & Hagen, 2008;Sala-i-Martin, 1996), although some differences may only become visible at a local level (e.g., Medeiros, 2014). Currently, we are observing a dynamic development of metropolitan areas forming growth poles (Kisiała et al., 2015;Novosák et al., 2017), and a stagnation in the development of peripheral areas (Bachtler & Gorzelak, 2007;Dall'erba & Le Gallo, 2008;Stanny, 2013). Notably, Stanny (2013) concentrates on the role of specific territorial factors in explaining CP effectiveness, where their diversity is most noticeable at a subregional level. Thus, incorporating these considerations into a single study not only makes it possible to answer the question of whether CP is effective, but also may help to identify the factors determining where and when it is possible to achieve the best results.
The success factors of the interventions indicated in the literature include the quality of government, which improves policy effectiveness (Rodríguez-Pose & Garcilazo, 2015;Rodríguez-Pose & Ketterer, 2020). Especially, countries with a high level of bureaucratic/administrative corruption (measured by the presence of human capital and good institutions: Becker et al., 2012Becker et al., , 2018Surubaru, 2017) tend to show a lower level of absorption capacity, which is another factor influencing the intervention effectiveness reported in the literature. These intangible factors, coupled with tangible issues such as social, technical and road infrastructure or access to a certain standard of goods and services, are seen as territorial capital resources (Fratesi & Perucca, 2014).
Notwithstanding the fact that the territorial capital concept has not been unambiguously defined and that its scale and value largely depend on the methods and measurement indicators adopted, some common features are mentioned by various authors. Van der Ploeg et al. (2008), Ventura et al. (2008) and Romão and Neuts (2017) agree that each area has its own specific resources, the use of which is the responsibility of the community and the local or regional economy. These resources are treated as common goods, and their utilization should contribute to territorial development. Morretta et al. (2020) add that the complementarity of the territorial capital elements is crucial for the more effective development and implementation of public policies, which can be key in the context of the absorption of European funds. Camagni (2008Camagni ( , 2009Camagni ( , 2017 proposes a classification of the characteristics of territorial capital, based on two main dimensions of rivalry (typology of goods) and materiality (degree of tangibility). In this conception, natural and cultural resources and public infrastructure are considered to be the least competitive but most material characteristics, and human capital and pecuniary externalities are the most competitive but least material characteristics. In the middle of this scale there are a number of characteristics that determine a territory's level of capital, and Tóth (2015) notes that measuring such a multidimensional concept of territorial capital requires the use of multivariate and in-depth analyses.
According to Fratesi and Perucca (2014), the relationship between territorial capital, CP intervention, and regional growth can be considered from short-, medium-and long-term perspectives. The short-term perspective is supposed to mediate the impact of the funds, but the effect may be different for each territory. In the medium-and long-terms, the intervention may be focused on the accumulation of territorial capital, which in will ultimately lead to regional growth. The specific level of territorial capital depends largely on the geographical location (Caldas et al., 2018;Fratesi & Wishlade, 2017;Gagliardi & Percoco, 2017;Le Gallo et al., 2011). Gagliardi and Percoco (2017) show that the greatest positive effect of CP appears in rural areas close to major cities. The benefits in these areas are also strengthened by a suburbanization process which creates an environment in which CP-based interventions can bring the highest rate of return (e.g., high population density and the affluence of local communities). However, in dispersed settlement in rural areas, this effect is negative though statistically insignificant. Crescenzi and Giua (2016) show that the positive CP effects are greater in regions with a more favourable socio-economic environment. Mieszkowski and Barbero (2021) come to similar conclusions when analysing Smart Specialisation projects founded by the European Structural and Investment Funds (ESIF) in Poland, and observed that the highest concentration of projects was around the largest cities, and shortages were seen in rural areas and smaller counties. This reveals a potential paradox of the EU's CP, whereby it seems to function better in regions that in fact have more favourable territorial resources.
Analyses aimed at evaluating CP effectiveness have usually been conducted with the help of advanced econometric models such as the HERMIN (Bradley, 2006) and QUEST (Ratto et al., 2009) models which focused on the effect of Structural Funds on basic macroeconomic indicators. The growing popularity of counterfactual methods has led to their increased use for assessing the impact of public policies on socio-economic development. Of the many counterfactual methods, several are used frequently and successfully in social sciences. Matching counterfactual methods (e.g., propensity score matching with the difference-in-differences approach; Bachtrögler et al., 2020) has been one of the most commonly implemented research methods in this field. However, due to certain limitations related to the binary dependent variable, modifications of the method have been sought in order to adapt it to the phenomena being analysed. The net effect has therefore been estimated in studies based on modified assumptions of counterfactual models such as the generalized propensity score (Becker et al., 2012;Mohl & Hagen, 2008), when estimating the effect of quantitative or continuous exposures on binary outcomes. Furthermore, various kinds of regression have also been applied. Notably, regression discontinuity design (RDD) has proved successful in assessing the outcome of EU grants for disadvantaged regions in member states (Becker et al., 2010;Percoco, 2017), as it exploits the discrete jump in the probability of EU transfer receipt at the 75% threshold for the identification of a disadvantaged region. However, the above-mentioned methods cannot be used when all of the units analysed are subject to intervention and differ in their initial development conditionsthat is, the territorial capital.
The great majority of studies are limited to measuring how intervention affects economic growth expressed in GDP (most often dynamically operationalized) (Becker et al., 2012;Bradley, 2006;Le Gallo et al., 2011;Gagliardi & Percoco, 2017;Sala-i-Martin, 1996), and a little less often to employment growth (Becker et al., 2010(Becker et al., , 2018. Bachtrögler et al. (2020) estimate the CP impact on the functioning of businesses in different territorial contexts by studying the dependent variable through three indicators: employment growth, added-value growth and productivity growth. The study by Caldas et al. (2018) is another interesting example where the authors concentrate on the impact of CP on Portuguese municipalities' performance and development. They use the dependent variable more extensively, using local development indicators such as population, purchasing power per capita, local economic growth, education, culture and sport, and indicators concerning the level of local debt. But at such a detailed territorial level of analysis, it would be impossible to account for the CP's impact on GDP.
In Poland, Wojtowicz (2019) researched the impact of CP on regional and local development using inter alia SPSM, and concluded that non-returnable grants to local governments improve the quality of basic infrastructure and stimulate changes in the structure of employment, but the question appears of whether it also contributes to long-term economic growthespecially in poorer areas. Wojtowicz's analysis was made on a poviat scale (the second-level unit of local government and administration in Poland), so mixing urban and rural areas in single research entity. Thus, the statement in Kozak's (2014) study that CP funding has so far favoured an improvement of quality of life rather than economic development is perhaps true, and illustrative of more demand-than supplyside effects.
The impact of territorial capital on Cohesion Policy in rural Polish areas 499

Variables characterizing the territorial capital of rural areas
In this study, the territorial capital of the administrative units is evaluated using empirical variables (Table 1) selected following a preliminary survey of the literature, and with respect to the proposed research methodology.
The choice was also based on conclusions drawn from long-term observations of the spatial diversity of socioeconomic development in Poland in the Rural Development Monitoring project (Rosner & Stanny, 2017). Specifically, it assumed that the variables had to fulfil all of the following conditions: (1) they should characterize the territorial uniqueness of rural Poland as comprehensively as possible in terms of the most important developmental resources (i.e., non-agricultural sector, agricultural sector, local finances, demographics, and social engagement; Stanny et al., 2018), but at the same time the range of measures examined should be limited to the most important ones; (2) statistical data should be gathered according to the same methodology for all the units, and should be available for a time period from around 2007 so as to reflect the status before the CP's impact; and (3) they should not be closely correlated with one another, but have a relatively high spatially diversity (as expressed by the coefficient of variation).

Outlays on rural areas under CP
Information on the outlays on projects completed under the CP between 2007 and 2018 was obtained from two government databases: KSI SIMIK for the 2007-13 financial framework, and SL2014 for the 2014-20 framework. 3 The dominant subject areas were aggregated based on a division into priority themes (Figure 1). Projects completed in 2007-18 with EU support and with convergence (cohesion) as their goal were worth approximately PLN 500 billion. Of this, 37% (approximately PLN 190 billion) was spent on projects in rural areas, which account for about 40% of Poland's population (Komorowski et al., 2021). The biggest part of the funding was spent on transport projects (PLN 70.5 billion, or 37% of projects completed in rural areas). Large amounts were also spent on projects involving technical facilities (PLN 43 billion), as well as enterprise, research and development (PLN 42 billion). Support was lower for social infrastructure (almost PLN 11 billion) and projects involving education and skills (almost PLN 6.7 billion). The least funding was allocated for projects involving cultural heritage and tourism (under PLN 5 billion), and environmental protection and digitization (approximately PLN 3.3 billion each).

Variables characterizing the social and economic conditions of rural development
The impact of CP in different municipality types (according to territorial capital) on social and economic changes in rural areas was measured using 38 variables. These were selected taking account of statistical, substantive and formal criteria. Their choice was the product of the availability of data which had to be methodologically comparable over a long period , be available for all municipalities, and be their best possible representation of the conditions of rural socio-economic development. Preliminary analysis (with a Kaiser-Meyer-Olkin measure < 0.5) showed that there was no validity in reducing the number of variables using factor analysis. The operational variables are presented in Appendix A in the supplemental data online. Though the source data for constructing the measures came from a number of public institutions (see Appendix A online), the final relative measures were obtained from the MROW database and supplemented with several variables developed on the basis of GUS data.

CP versus the Common Agricultural Policy (CAP)
Another public policy aimed at support for rural areas is the CAP, but which could not be included in the modelling due to the lack of available data (despite multiple requests), even after the formal deadline of the project. Nevertheless, we decided to continue the modelling, as the goals and scope of the CAP were different from the goals and scope of CP. The separate nature of these two policies can be confirmed mainly by the multiplicity and availability of relevant scientific studies on the impact of the CAP on rural development (e.g., Esposti, 2017;Michalek et al., 2020), whereas for the CP, studies are more general and cover whole countries. Moreover, the CAP was designed to serve price and market interventions, direct payments, environmental purposes and rural development (Alons, 2017), and its unquestioned contribution to economic cohesion can be seen as a by-product of its specific objectives (Hansen & Herrmann, 2012). While the CP was intended to operate in areas of general growth (such as business competitiveness, transport, environment and energy, digitalization, education, health or the labour market), both in urban and rural areas, the CAP's scope is to improve the competitiveness of the agricultural sector and to enhance the environmental conditions and quality of life in rural areas mainly in relation to agricultural activities (Drygas & Nurzyńska, 2021). A detailed analysis of CAP and CP measures for the financial perspective of 2021-27 proves that the CP is intended to solve socio-economic problems in rural areas, while the impact of the CAP might be narrowed down to production and environmental issues in the agricultural sector (Wasilewski et al., 2021). However, as only 13 of the 49 support areas examined in the study are scheduled in both policies, their complementarity is limited.
The divergence is also visible in CP and CAP expenditure. Under the first pillar of the CAP, since Poland's accession to the EU, funds have been spent within the framework of direct support and on mechanisms for the common organization of agricultural markets. The direct payments amounted to PLN 172.1 billion in the period of 2004-18. Market mechanisms concern fruit and vegetable producer groups (which have been granted PLN 7.9 billion since accession to the EU), and the organization of other agricultural markets (supported with PLN 443.3 million from September 2017 to the end of 2018).
The Rural Development Programme is an instrument under the second pillar of the CAP to promote sustainable development in rural areas. RDP 2007-13 was completed on 31 December 2015, and PLN 74.3 billion was spent (together with local government and other public funds) within its framework. Most funds (PLN 33.4 billion) were allocated to the implementation of Axis 1, which is dedicated to improving the competitiveness of the agricultural and forestry sectors. Outlays on Axis 2 (improving the environment and rural areas) amounted to PLN 22 billion, and measures under Axis 3 (quality of life in rural areas and the diversification of the rural economy) received PLN 14.5 billion. A smaller amount, PLN 3.4 billion, was allocated to the implementation of Axis 4, LEADER, and PLN 1 billion was spent on technical assistance.
The 2014-20 RDP is currently being implemented. The programme budget is around 60 billion PLN, of which the largest amount is planned for investments in fixed assets (26.5%), the development of farms and business activities (16.9%), and payments for areas with natural or other specific constraints (14.7%). Approximately PLN 4.7 billion (8.1%) is planned for measures in the area of basic services and village renewal in rural areas, while the LEADER initiative will be supported with approximately PLN 3.2 billion (5.4%). By the end of 2018, payments of PLN 16.5 billion had been made.
To conclude, the total funds spent on rural areas under the CAP amounted to about PLN 270 billion, 4 while under the CP they were assigned approximately PLN 190 billion. However, over PLN 180 billion of the CAP involved direct payments for farmers. Only PLN 90 billion was allocated to development processes, most of which concerned activities related to agriculture (rather than quality of life in rural areas and diversification or local cooperation under the LEADER initiative) (Figure 2). Although the CAP has had a significant impact on rural areas in Poland, its goals diverge from the CP and its influence on the assumed impact indictors is very limited. Accordingly, its impact may be understood as a variance of indicators not explained by CP investments, although in future studies it would be beneficial to integrate both programmes if suitable data was available.

STRATIFIED PROPENSITY SCORE MATCHING (SPSM)
As macroeconomic models have serious theoretical limitations such as simplifying the functioning of a given economy or assuming only positive changes resulting from EU investments (Gorzelak, 2009), another approach to impact evaluation was developed. This involves experiments and quasi-experiments, mostly matching methods which transform non-experimental data into quasi-experimental data, in order to create counterfactual situations (Strawiński, 2008). The purpose is to overcome the effect of non-random selection on the experimental and control groups, and the subsequent counterfactual analysis is a comparison between what actually happened and what would have happened without intervention (White, 2006). Practice shows that the impact of EU funds is widely measured using a quasi-experimental approach based on propensity score matching (Konarski & Kotnarowski, 2007), despite its limitations resulting from e.g., the use of logistic regression, intangible net effects (Wojtowicz et al., 2010), or insufficient sample size (Trzciński, 2009). However, the above methods would not provide adequate results for this study, and as all of the local government units in Poland have benefited from CP funds, it was impossible to define the experimental and control groups that are usually compared during traditional counterfactual analysis. The main research method applied in this study was therefore SPSM (Wojtowicz & Widła-Domaradzki, 2017), which allowed us to estimate the net effect for units, all of which were subject to intervention (i.e., were beneficiaries of CP funds) to a greater or lesser extent.
The main stages of the analysis were hierarchical clustering and fuzzy clustering, leading to the calculation of propensity scores (Widła-Domaradzki, 2017). The hierarchical clustering of local government units was based on Ward's algorithm (Ward, 1963), using the initial values of the dependent variables (before the intervention). Ward's algorithm analyses the error sum of squares at each step as follows: where x i is the score of the ith unit; and n is the sample size. The algorithm is repeated until a sufficient number of clusters is obtained. Each iteration is defined as follows: where p n−1 is the smaller of the two clusters; and q n−1 is the larger. As a result, every unit is assigned a discrete cluster membership value. Hierarchical clustering was followed by fuzzy clustering of the units, resulting in the calculation of cluster membership probability which was regarded as a propensity score (based on discriminant analysis). Having obtained the propensity scores, one may proceed to order and pair the most similar units according to the propensity score for each cluster. This allows each pair of units to be defined as a counterfactual situation. The next step was estimating the values of the net effects for every dependent variable concerning the entire sample and within the clusters, as differences in the values of changes within the pairs of units. The interpretation of the value obtained is based on the assumption that municipalities within a single counterfactual situation should achieve a parameter change within a similar range. If they achieve significant differences in values, it means that additional external events affecting a given unit's development have taken place.
The net effects were later used to assess particular EU funds or specific interventions, as well as for regression analysis where the determinants of the changes observed were identified. An observation included in the regression models was a pair of examined units (a counterfactual  Agata Mróz et al.

situation)
, not a single unit, which allowed the actual determinants of any net effects to be identified.

Clustering based on Ward's algorithm
Based on the set of variables describing rural areas' territorial capital, the municipalities were divided into five groups, that is, unit types that were relatively homogeneous according to the criteria adopted, and which are maximally different from one another. The territorial distribution of these types is shown in Figure 3. The units with the most favourable territorial capital characteristics (Table 2) are type 5 municipalities, whose core is in the vicinity of big cities or regional capitals. However, the range of the influence of such neighbourhoods varies. For example, in eastern Poland it is noticeably weaker compared with the rest of the country, and it is also related to the size of the urban unit, extending farthest in the case of the Warsaw conurbation, Poznań, as well as in the south around Wrocław, Katowice and Kraków. The strength of this influence varies in different directions which is linked to the layout of the transport network and also the gravitation of transport links towards other big cities. Consequently, these may be seen as wealthy (I3) multifunctional municipalities (with a reduced agricultural function I2), and their strength is high social engagement (I5).
The above-average value of territorial capital in type 1 municipalities is also noteworthy. This indicates the multifunctional character of rural areas (I4) that are demographically relatively young. The high evaluation of the development of non-agricultural functions (I1enterprise) coupled with a relatively good evaluation of the agricultural sector (I2) shows the competitive advantage of these municipalities over other rural areas. Geographically, municipalities of this type stretch longitudinally from Gdańsk to Wrocław, through western Poland.
Type 2 municipalities are referred as post-PGR areas, due to the fact that state-owned farms (PGRs) were concentrated there until 1990. 5 This is a unique formation mainly occurring in northern Poland. The restructuring of PGRs gave the current structure of the agricultural sector the characteristics of large-scale farming, based on hired labour. In these areas, enterprise (I1) is close to the median for all Polish municipalities. The weak link in the evaluation of territorial capital in this type of municipality is their residents' poor social engagement (I5) and the municipalities' relatively low financial wealth (I3).
Type 3 municipalities are characterized by a description of territorial capital that is closest to the average. These are areas with an unfavourable evaluation of the agricultural sector, characterized by agrarian fragmentation (I2) and multi-income farming families dealing with daily commutes to nearby large and medium-sized cities. In this type of municipality, agriculture is losing its importance as a source of income, and the development of a multifunctional economy is not connected with land concentration but, very often, with the decline of agriculture.
The least favourable territorial capital is to be found in rural areas in central and eastern Poland. These are type 4 municipalities whose characteristic features include a very high proportion of agriculture in the rural economy, coupled with a very traditional (peasant) approach to farming. Throughout the twentieth century, the poorly developed non-agricultural labour market caused the migration of young people to cities, which led to an unfavourable demographic structure developing in these areas, with a very high proportion of older people (I4). A low level of enterprise in these municipalities translates not only into their lowest affluence, but also into a relatively low quality of social capital (I5).

Effects of CP investments in rural areas
Within the municipality types described above, the SPSM enabled the identification of pairs of units that were the most alike in terms of the value of territorial capital criteria before the intervention began. Next, the differences in the values of changes (in the years 2007-18, or other widest possible range) of the social and economic variables in a given area (explained in the regression models) were calculated.
The effects of CP intervention were estimated based on the construction of several hundred regression models for 38 indicators in the five municipality types. The value of the coefficient of determination R 2 (informing which part of the response variable's changes is explained by changes in the explanatory variable) was obtained for each of those models. Depending on the type of municipality and the specific choice of explanatory and response variables (i.e., the specification of the model), anything from a few percent up to approximately 80% in the variation of socio-economic indicators can be attributed to CP-supported investments. Due to the generally low values of the coefficient, a value of at least 20% was assumed to be the limit for the existence of influence. Nevertheless, even with this assumption, it was not possible to prove the CP's impact for any indicator in municipalities with adverse demographic processes, a weak financial situation of local governments and a high share of agriculture in their local economy, which were located mainly in central and eastern Poland (type 4), as the highest value of R 2 in this group was 18%. The power of the influence of CP outlays on changes in the social and economic conditions of rural development are presented in Appendix A in the supplemental data online, while a range of conditions with the greatest impact on social and then economic aspects of local development is outlined below.

Impact on social aspects of local development
It was possible to verify the greatest impact of support from the CP in suburban municipalities (type 5) with the highest share of non-agricultural entities on social indicators such as a higher natural increase per 1000 inhabitants and population change, an increase in the The impact of territorial capital on Cohesion Policy in rural Polish areas 503 feminization rate, and a decrease in the intensity of commuting, and the percentage of families covered by the social welfare system (Figure 4).
In municipalities with a relatively high level of social and economic development, with favourable parameters of both the non-agricultural and agricultural sectors  (type 1), one can see the impact of CP support in a decrease in the intensity of commuting, a higher natural increase per 1000 inhabitants and a higher change in population. The CP's impact on demographic indicators was therefore particularly noticeable in municipalities located close to larger cities, which may indicate the diffusion of development processes into suburban areas. On the other hand, the demographic changes seen here result from location advantages around cities, as economically active younger people often choose to live in the suburbs. It has not been possible to verify the CP's impact on the rate of attractiveness for migration or the intensity of commuting in units of types 3 and 4, which may be due to insufficient and dispersed support for the development of enterprises in peripheral rural areas. The econometric analyses did not make it possible to verify CP's impact on educational indicators; for example, on the gross enrolment rate at the secondary school level, or the results of primary and secondary school exams. The spatial diversification of exam results is rather related to cultural conditions, where better grades are obtained in the south. In addition, the effects of education support are not immediately visible, as they are shifted in time, and also in space because of migration.

Impact on economic aspects of local development
It was possible to verify the impact of CP support in suburban municipalities with the highest share of non-agricultural entities (type 5) on the increasing share of these entities in the total number of businesses, and therefore also higher non-agricultural tax incomes. In municipalities with relatively favourable agricultural characteristics where large-scale farms predominate (type 2), the impact of CP support can be seen in a decrease in the number of people in the Agricultural Social Insurance Fund (KRUS), and an increase in the number of personal income tax (PIT) and corporate income tax (CIT) taxpayers per 1000 inhabitants. In type 1 and 2 municipalities characterized by relatively high territorial capital, one can see the impact of CP support in an increase in the share of non-agricultural entities. A noticeable relatively weaker dependence was observed in type 3 municipalities characterized by fragmented agriculture and a tradition of commuting by the 'bi-occupational' population (e.g., part-time farmers who also work in industry).
The impact of support from the CP on economic indicators is presented in Figure 5.
The SPSM analysis did not make it possible to verify the CP's impact on the intensity of unemployment (including long-term unemployment), or the number of businesses in the official register (REGON) per 1000 people of working age in any type of municipality (for variables, see Appendix A in the supplemental data online). This may be due to a number of reasons, including labour migration to cities (which causes a shifting of the effects of intervention in space), or a failure to adapt interventions to the needs of the long-term unemployed and the excluded. Other reasons may be a maintenance of unemployed status by people working in the shadow economy, and the lack of impact of interventions on official statistical indicators.
It was also impossible to verify the CP's impact on some other indicators, including the average travelling time to the province and powiat/county capital, the percentage of villages in a municipality connected by public transport, and the percentage of children attending kindergarten (for the variables, see Appendix A in the supplemental data online). This may be due to the fact that the highest outlays under the CP went to the construction of motorways and expressways, while local roads were also developed under the Rural Development Programme and other funds. The construction of new roads and the provision of technical infrastructure improves the quality of life. However, the effects vary depending on the level of local development. Better roads may encourage new residents to settle in more attractive areas, which after time, causes an increase in traffic, congestion and the need to re-expand the infrastructure, both technical (e.g., watersupply, sewage management and roads) and social (e.g., kindergartens and schools). Furthermore, in areas with an underdeveloped labour market it may make people commute longer distances to places offering higher wages, and so may have no economic effect at all in the peripheries.

DISCUSSION AND CONCLUSIONS
The analysis evaluated the CP impact on the development of rural areas in Poland taking into account their spatial diversification, and offering a new perspective in terms of both rural studies' methodology and territorial approach. Certain limitations resulted from three specific areas: (1) the difficulty in accessing statistical data describing the social and economic conditions of rural development over long periods at the local level; (2) the operationalization of territorial capital, mainly in terms of access to variables characterizing its tangible and intangible aspects in municipalities; and (3) difficulties in distinguishing CP-based intervention in the rural parts of urban-rural municipalities. However, the study makes a significant contribution to the discussion on the effectiveness of CP in rural areas with different territorial potential.
The effects of the CP in Polish rural areas turned out to be very diverse, depending on the territorial capital type of the local units. It was possible to verify the greatest positive impact of CP support on social and economic conditions in municipalities with relatively favourable territorial capital, specifically in multifunctional, urbanized municipalities with a reduced agricultural function. However, it proved impossible to demonstrate the CP's impact on rural development in municipalities with adverse demographic processes, a weak financial situation of local government, and with a high proportion of agriculture in the local economy. Nevertheless, a weak impact of CP support was observed in municipalities with a relatively unfavourable evaluation of the agricultural sector (i.e., fragmented agriculture) and a tradition of commuting by the bi-occupational (farmer-worker) population, who The impact of territorial capital on Cohesion Policy in rural Polish areas 505 supplement their farm income in the city. Hence the hypothesis of a more significant impact of CP on economic rather than social conditions of rural development was not supported. Although the indicators for social conditions to a greater extent determined the effects observed, we are aware that the complexity of intervention related to social aspects is decidedly more difficult to measure. As Rodríguez-Pose and Fratesi (2004) have noted, its effect is often delayed only becoming noticeable some time after the intervention, and in the meantime, other interventions may follow. The impact on social conditions was only noted in multifunctional, urbanized municipalities. The study has confirmed that the structure of the local economy is the main factor determining the impact of intervention under the CP, wherein, the more advanced the development of non-agricultural functions (enterprise), the multifunctional character of a rural area, and the reduction of the agricultural function, the greater is the effectiveness of CP. It was also confirmed that in rural areas, important factors determining intervention effectiveness include a municipality's financial resources (the affluence of the local government) and (as literature  has already suggested; Becker et al., 2012Becker et al., , 2018 human capital, which together with a good institutional base, determine an area's absorptive capacity. These are most often the characteristics of wealthy, densely populated suburban municipalities in the vicinity of major cities, which means that the third hypothesis, indicating the importance of location advantages (Crescenzi & Giua, 2016;Fratesi & Perucca, 2014;Gagliardi & Percoco, 2017) was also confirmed. 6 Insofar as the level to which the development of Polish regions is approaching the EU average (Monfort, 2020); when we look at the national scale, the differences between areas with low and high potential territorial capital are deepening. This division has a functional axis with urban areas as the centre and rural areas as the periphery, as well as a geographical axis where more favourable changes are seen in western rather than in eastern Poland. For the latter condition, the foundation is historical and reflects a powerful path dependence (Churski et al., 2020) which lies beyond the scope of this analysis. The low effectiveness of CP is highly noticeable in municipalities with less favourable territorial capital. In such areas, there is a need for a continual building of human and social capital as the basis of intangible local resources, including relational capital and the capacity for good government in Poland's rural areas. Alongside investing in technical and social infrastructure, we should also (perhaps above all) be building and developing intangible capital, which in the case examined has been a neglected aspect in the past. Nevertheless, this should be a key object of public policy intervention, including CP implementation in successive programme periods.
Increasing absorptive capacity in areas with less favourable territorial capital is a challenge for further experiments and research, together with how to design interventions so that future public measures in rural areas in Poland and similar areas around the world are more effective in stimulating convergence processes. Inherently, evaluations need to move towards a more complex analysis of strategic impacts, taking into account all the possible determinants of the impact. Compared with previous analyses, this study integrated the impact with a holistic focus on the development of rural areas, but there is still a need to further integrate its factors.

ACKNOWLEDGEMENTS
The study used data from the Rural Development Monitoring (MROW) project carried out jointly by the European Fund for the Development of Polish Villages and the Institute of Rural and Agricultural Development, Polish Academy of Sciences, for which the authors are grateful.

DISCLOSURE STATEMENT
No potential conflict of interest was reported by the authors.

NOTES
1. The municipality (gmina) is a local administrative unit (former NUTS-5). Poland has three types of administrative municipality: urban, urban-rural and rural. On average, a rural or urban-rural municipality has a population of about 9000 inhabitants, and an area of 140 km 2 . 2. The research methodology takes into account the key economic and non-economic determinants of rural development which are essential to address the features of rural areas in Poland. These factors include the de-agrarianization of the local economy, the characteristics of agricultural and non-agricultural sectors, the spatial accessibility of municipalities, local public finances, the labour market, demographic issues, education, social activities, wealth, and the living conditions of the local community. 3. Due to some projects still being implemented as part of the 2014-20 framework and the time needed for intervention effects to be observable, the analysis only covers completed projects. 4. PLN 1 ¼ €0.23 (2018 average). 5. The Państwowe Gospodarstwo Rolne (PGR), or state agricultural farm, was a form of collective farming in the People's Republic of Poland, similar to the Soviet sovkhoz and East German Volkseigenes Gut. 6. However, the scale of these differences is something of a surprise, as the effects of intervention are barely noticeable in the most backward areas. It needs emphasizing that other studies using the SPSM method (Wojtowicz & Widła-Domaradzki, 2017;Wolański et al., 2018) made it possible to identify a much wider effect of EU intervention in Poland, so it would be wrong to equate the lack of effects with the ineffectiveness of the method.