Urban development and expenditure efficiency in the 2000–2006 regional operational program of Sardinia

Land Use Policy 28 (2011) 472–485

Contents lists available at ScienceDirect

Land Use Policy journal homepage: www.elsevier.com/locate/landusepol

Urban development and expenditure efficiency in the 2000–2006 regional operational program of Sardinia Corrado Zoppi ∗ , Sabrina Lai 1 Dipartimento di Ingegneria del Territorio, University of Cagliari, Via Marengo 2, 09123 Cagliari, Italy

a r t i c l e

i n f o

Article history: Received 14 January 2010 Received in revised form 27 September 2010 Accepted 1 October 2010 Keywords: Territorial cohesion Regional operational programs Geographic information systems Multinomial logit models

a b s t r a c t The Sardinian regional operational program 2007–2013 concerning the European Regional Development Fund (ERDF) (2007–2013 ROP-ERDF) respects the rules of the ERDF on the investments for territorial cohesion since it promotes their regional geographic concentration. These investments are evenly shared between large- and medium-sized urban areas, and disadvantaged zones. This paper analyzes the investment policies implemented by the Sardinian Region through the 2000–2006 ERDF based part of the Regional Operational Program (2000–2006 ROP-ERDF), in order to assess their effectiveness, in terms of expense efficiency, for urban areas and disadvantaged zones. The assessment of the expenditure efficiency of the 2000–2006 ROP-ERDF is very important to address the policies of the 2007–2013 ROP-ERDF in terms of geographic concentration. The essay analyzes the results of the 2000–2006 ROP-ERDF, with reference to the expenditure efficiency of Sardinian cities, represented through a geographic information system, by means of a multinomial logit model. The essay proposes an analytical and interpretive approach which could be easily exported to other public planning processes, in order to define policies for territorial cohesion. © 2010 Elsevier Ltd. All rights reserved.

1. Introduction The principle of territorial cohesion was officially introduced in the European Union Treaty by the Treaty of Amsterdam (1997), which amends the European Union Treaty by adding Article 7d as follows: “Without prejudice to Articles 77, 90 and 92, and given the place occupied by services of general economic interest in the shared values of the Union as well as their role in promoting social and territorial cohesion, the Community and the Member States, each within their respective powers and within the scope of application of this Treaty, shall take care that such services operate on the basis of principles and conditions which enable them to fulfill their missions” (art. 2, par. 8). Afterwards, this concept was reconsidered and clarified in other important documents such as the European Spatial Development Perspective (1999), the Leipzig Charter on Sustainable European Cities (2007) and the working document of the Directorate General for Regional Policy of the Commission of the European Communities “Fostering the urban dimension. Analysis of the Operational Programs co-financed by the European Regional Development Fund (2007–2013)” issued in November 2008. These documents are

∗ Corresponding author. Tel.: +39 070 6755216; fax: +39 070 6755215. E-mail addresses: [email protected] (C. Zoppi), [email protected] (S. Lai). 1 Tel.: +39 070 6755200; fax: +39 070 6755215. 0264-8377/$ – see front matter © 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.landusepol.2010.10.001

conceptually articulated and put in evidence, with reference to territorial cohesion, a careful integration effort with respect to regional and urban spatial planning traditions and practices, which are very different from each other amongst the EU countries. Territorial cohesion is recognized as an important informative principle of the EU 2007–2013 cohesion policy, based on the Structural Funds. Regulation no. 1080/2006/EC, concerning the European ERDF, states that “in the case of action involving sustainable urban development as referred to in Article 37(4)(a) of Regulation (EC) no. 1083/2006, the ERDF may, where appropriate, support the development of participative, integrated and sustainable strategies to tackle the high concentration of economic, environmental and social problems affecting urban areas. These strategies shall promote sustainable urban development through activities such as: strengthening economic growth, the rehabilitation of the physical environment, brownfield redevelopment, the preservation and development of natural and cultural heritage, the promotion of entrepreneurship, local employment and community development, and the provision of services to the population taking account of changing demographic structures.” (art. 8) Moreover, the questions of territorial cohesion are stressed in Article 10, which establishes that the ERDF can finance investments for the urban, economic and social development of “areas facing geographical and natural handicaps as referred to in point (f) of Article 52 of Regulation (EC) no. 1083/2006,” which, in the Sardinian case, are identified with mountain areas, and zones with low (less than

C. Zoppi, S. Lai / Land Use Policy 28 (2011) 472–485

50 inhabitants per km2 ) and very low (less than 8) demographic density. The 2007–2013 ROP-ERDF respects the rules of the ERDF on the investments for territorial cohesion since it promotes their regional geographic concentration (Article 37(3) of Regulation (EC) no. 1083/2006). These investments are evenly shared between large- and medium-sized urban areas, and disadvantaged zones, as defined above. This paper analyzes the investment policies implemented by the Sardinian Region through the 2000–2006 ROP-ERDF, in order to assess their effectiveness, in terms of expense efficiency, for urban areas and disadvantaged zones. According to this interpretation of the concentration principle, which entails also thematic and financial issues, urban areas will benefit from funds mostly coming from the implementation of Axis V-Urban Development and Axis VI-Competition of the 2007–2013 ROP-ERDF, while disadvantaged zones will be mainly affected by the implementation of the first four axes (“Information society,” “Inclusion, social services, education and legality,” “Energy,” “Environment, natural and cultural attractions, tourism”) (2007–2013 ROP-ERDF, pp. 161 and ff.). It is therefore evident that the role territorial dimension plays is decisive for the effectiveness of the 2007–2013 ROP-ERDF, and, ultimately, for the unified regional programming activity for the period 2007–2013, which is strictly connected to the 2007–2013 ROPERDF. The most important regional program for the 2007–2013 period is the 2007–2013 Regional Implementation Program of the National Fund for the Underutilized Areas, established by the Deliberation of the Sardinian Regional Government (DRG) no. 71/47 of December 16, 2008. This program integrates the 2007–2013 European programs for the Sardinian region in order to build-up the Sardinian Regional Unified Programming Document, established by the DRG no. 64/9 of November 18, 2008. In this regional programming framework, the analytical assessment of the implementation of the investment policy for the period 2000–2006 with regard to the territorial dimension of the ERDF funds is certainly of paramount importance for the investment policies of the 2007–2013 regional programming. This essay offers a contribution for this assessment, since it analyzes and assesses expenditure efficiency with reference to its territorial characteristics. Analysis and assessment are implemented by assuming the cities as the basic spatial units. The essay has three parts. In the first part, a discussion on expenditure efficiency and its related characteristics, and a qualitative analysis of how the funds were spent, are proposed. A qualitative analysis of how the funds were spent concludes. In the second one, a territorial taxonomy of the cities is developed with reference to the expenditure efficiency, represented through a geographic information system (GIS). Thirdly, the expenditure efficiency phenomenon is studied by means of a multinomial logit model (MNL model), whereby correlations between the variables contained in the spatial database built in the previous section are identified and analyzed. In the conclusions, it is put in evidence and discussed how and why the proposed approach can be considered a first step toward a regional programming policy mainly based on a vision inspired by integrated urban development rather than by sectoral objectives, which is consistent with the European Commission vision on fostering the urban dimension (Commission of the European Communities, 2008, pp. 3–7).

2. Expenditure efficiency and its determinants In this paper, expenditure efficiency of a city is defined with reference to the capability of spending the projected ROP 2000–2006 investment within the established time. This broad and apparently

473

rather simplistic definition is very effective in capturing a city’s attitude toward defining and implementing plans and programs within the ROP. This assumption is motivated as follows. Under the “N + 2” rule, set out in Article 31 of European Union Council Regulation no. 1260/1999 as a financial control designed to encourage sound management of the programs and avoid endloading of activity, de-commitment of any funds not yet spent by the end of the second year following the year to which they were allocated is required. Thus, the very fact that local administrations succeed in avoiding end-loading, and in organizing their investment policy harmoniously along the whole ROP lifetime, is regarded as very important by the European Union. What can be observed in the case of Sardinia is that the final phase of the ROP was characterized by a tendency to spend in any case, provided that, by doing so, cities could avoid being held responsible for the loss of conspicuous financial resources. In this context, cities that successfully avoided the 2009 final spending rush could be comparatively assumed to be more efficient in programming and implementing their ROP investment. Recalling that, under the “N + 2” rule, the expenditure of the 2000–2006 ROP had to be entirely realized and the official statements of account had to be made available to the European Commission by June 30, 2009, we assume that the amount of investment spent within nine months from the final term (September 2008) is a good measure of the efficiency of cities in implementing their targeted ROP investment. We consider expenditure efficiency as connected to the following attributes of cities. The number of residents is an important indicator of the economic, social and political relevance of a city; the residential density accounts for the relationship between the territorial organization of the residential fabric and the administrative area of the city, since (everything else being equal) the higher the residential density, the higher the land consumption, which also implies a preference for renewal/reuse building policies with respect to new residential developments. Moreover, the more consumed the land, the greater the attention to be devoted to natural resource protection. The typological taxonomy of the cities is related to urban development policies and to policies for disadvantaged zones, which found the approach of the 2007–2013 ROP-ERDF to geographic concentration of the investments, described in the introduction. This characteristic makes it possible to compare, qualitatively and quantitatively, the performances of disadvantaged zones and of urban areas, in terms of expenditure efficiency, with reference to the seven-year period 2000–2006. In other words, it is possible to assess whether the investment policies for territorial cohesion based on geographic concentration for the 2007–2013 ROP-ERDF can rely upon good and balanced results in terms of expenditure efficiency with reference to the outcomes of the 2000–2006 ROPERDF. The relation between expenditure efficiency and the fact that a city is coastal or non-coastal allows to detect if it is influenced by this characteristic. This is important because coastal cities in Sardinia, starting from 2004, had to cope, in projecting their urban planning policies, with the implementation code of the Sardinian Regional Landscape Plan (RLP, approved by the DRG no. 36/7 of September 5, 2006), which introduced severe limits and a very strict regulation concerning new developments and territorial transformations for the cities which are part of the coastal zone as defined by the RLP cartography (Zoppi, 2008). Projected investment of a city from the 2000–2006 ROP-ERDF is related to the relative importance given by the program to that city with respect to the regional framework, and allows to assess whether a city behaved satisfactorily, in terms of expenditure efficiency, having regard to the amount of money it was granted by the program.

474

C. Zoppi, S. Lai / Land Use Policy 28 (2011) 472–485

The paragraph below contains qualitative information analysis on how the ROP-ERDF funds were spent. 2.1. A qualitative analysis on how the expenditure of ROP-ERDF funds during the 2000–2006 programming period Up to 2006, Sardinia was an Objective One region, as its gross domestic product (GDP) per capita in 1996 was approximately 73 percent of the average GDP in the European Union (Hospers, 2003:632), therefore below the threshold of eligibility (75 percent). As such, Sardinia was included in the highest priority area designated for European aid, where funds aimed at supporting ‘economic growth and sustainable development through investment in people and in physical capital’ (CEC, 2004) by eradicating the roots of regional disparities, identified by the European Commission in: lack of infrastructure and human capital, inadequate innovative capacity and difficulty in sustaining economic growth. Sardinian strategy and regional priorities for action for the delivery of the funds were set out in its 2000–2006 ROP, and classed into seven axes as follows: (I) – natural resources; (II) – cultural resources; (III) – human resources; (IV) – local development systems; (V) – cities; (VI) – networks and service hubs; (VII) – technical assistance. Each of the axes was divided into a number of measures, further subdivided into actions. The total expenditure programmed by the ROP in its latest version2 equaled D 4,258,555,040.00 (Regione Autonoma della Sardegna, 2007a,b, Annex 3 – financial tables); approximately a 50 percent of the funds came from the European Union, a 35 percent from the Italian government, and the remaining 15 percent from the regional government. Unlike what happens in the current 2007–2013 programming period, where a ROP has to be approved for each source of funding, in the 2000–2006 period ROPs set out a strategic framework for projects to be funded through the cohesion policy, irrespective of the type of fund (apart from the ERDF, these were the European Social Fund, ESF, the European Agricultural Guidance and Guarantee Fund, EAGGF, and the Financial Instrument for Fisheries Guidance, FIFG). Nearly a half of the total investment programmed by the Sardinian 2000–2006 ROP concerned measures co-funded by the ERDF, for which a total amount of D 2,600,980,000.00 was reserved (Regione Autonoma della Sardegna, 2007a,b, Annex 3 – financial tables); a half of these, as previously stated, came from the European Union. Table 1 lists all of the measures co-funded by the ERDF, together with the total programmed investment per measure and the share of the total cost funded by the ERDF. Measures contained in the 2000–2006 ROP are described in detail in a separate document, the so-called “Complement Programme” (CP). For each measure, the CP gives indications on the administrative procedure to be followed for its implementation, which, in the vast majority of cases, was a call for proposal issued by the public organization in charge of the implementation of a certain measure, usually a Department of the Regional Administration of Sardinia. Depending on the measure, project proposals could be submitted by either public administrations (such as, for instance, local governments), or by private organizations and individuals. For each measure, eligibility criteria and selection criteria on which to evaluate project proposals were also identified in the CP. A close analysis of the CP shows that only public administrations could apply to funds concerning eleven out of the 22 measures co-funded by the ERDF, hence a total of D 1,755,735,560.00 was

2

Approved by the Commission Decision C (2007) 1991 April 30, 2007.

reserved for public projects. As far as the remaining eleven measures are concerned, Table 1 breaks down each measure into actions and puts in evidence whether final beneficiaries were private or public organizations. This shows that funds were mostly spent on public services and infrastructures, with only a relatively small share of funds reserved for the private sector under the European provisions on State aid. The bulk of ERDF funds was therefore allocated to public projects, and mostly to public infrastructures.

3. Representation of the expenditure efficiency by means of a GIS This section describes the geographic information system (GIS) developed to assess Sardinian municipalities’ performances with regard to expense efficiency. The aim of this paper is to investigate whether, and to what extent, the ability of Sardinian cities and towns to spend timely those funds they had been granted by the 2000–2006 ROP was correlated to a series of variables introduced in the previous section (that is, population, residential density, taxonomy of the cities, proximity to the coastline). In order to perform such analysis by means of the MNL model, a descriptive table was needed. Such table had to have as many rows as Sardinian municipalities are, and as many fields as relevant attributes are, one of the fields necessarily being expenditure efficiency, already defined in the previous section, and analyzed only in reference to projects co-financed by the ERDF during the programming period 2000–2006. Data from various sources and in various formats were collected; some data (e.g. those pertaining to demography and levels of expenditure) were available as spreadsheets; some others (e.g. coastline) were available as spatial databases. Moreover, integration of available (both spatial and non-spatial) information was required to develop new knowledge and obtain new layers of either spatial or non-spatial information, which called for a GIS-based analysis. As far as the geographic description of the municipalities is concerned, only one shapefile was used to implement the GIS. This shapefile, produced by the Regional Administration of Sardinia on the basis of their digital cartography,3 contains 377 polygons, one for each municipality; each polygon in the map corresponds to the land area included within a municipality’s administrative boundaries. The attribute table associated with this polygonal theme consisted therefore of 377 rows (records, as many as the municipalities are) and four columns (fields). The four fields in the original table contained only some basic information about each municipality: land area, length of the boundary, name of the municipality, and an alphanumeric code which identifies uniquely each municipality in the Italian Census system (ISTAT). This table was therefore completed by means of a series of “join” procedures4 leading to the addition of a number of fields, whose labels are as follows: (i) POP2001, (ii) POP2007, (iii) POPCH0107, (iv) DENS2001, (v) DENS2007, (vi) COAST, (vii) TYP CITY, (viii) PROJ INV, (ix) EXPEND, and (x) EFFIC.

3 Sardinian Regional Digital Cartography was produced between 1994 and 2000. It is a CAD-based cartography which covers the whole regional territory (approximately 24,000 km2 ) and has a scale of 1:10,000. Coordinates are given with reference to the Italian national grid (also known as Rome 40 – Monte Mario) (http://www.sardegnaterritorio.it/j/v/241?s=15891&v=2&c=1938&t=1 (accessed July 19.07.10)). 4 Chrisman (2002, p. 133) defines “join” as “a procedure that attaches values from a database table to another table based on matching a foreign key to its primary instance.”

C. Zoppi, S. Lai / Land Use Policy 28 (2011) 472–485

475

Table 1 Measures of the Sardinian ROP 2000–2006 co-funded by the ERDF. Total cost, ERDF share and final beneficiaries (Regione Autonoma della Sardegna, 2007a,b, Annex 3 – financial tables). Measure

ERDF co-funding [D ]

Beneficiariesa 100% PUB 100% PUB 100% PUB

1.1 1.3 1.4

Integrated Water Cycle Soil Conservation Integrated Waste Management and Remediation of Polluted Soils

527,899,628 275,606,000 74,388,000

263,949,814 137,803,000 37,194,000

1.5

Sardinian Natura 2000 Network

49,327,000

24,663,500

1.6

Energy

21,883,000

10,941,500

1.7

Environmental Monitoring System Archaeology, Religion and Museums Networks, Restoration of Historical City Centers for Culture- and Tourism-led Activities Industrial Archaeology

28,511,000

14,255,500

149,274,440

74,637,220

49,576,932

24,788,466

Structures and Services for Cultural and Entertainment Activities

156,816,000

78,408,000

62,752,000

31,376,000

70,191,130

35,095,565

2.1 2.2 2.3 3.12

Structures to Incentive Education; Employment Services Research and Technological Development in Enterprises and Local Contexts

1.5.a: PUB 1.5.b: PUB 1.5.c: PRI 1.6.a: PUB 1.6.b: PRI 100% PUB 2.1.a: PUB 2.1.b: PUB 2.1.c: PRI 100% PUB 2.3.a: PUB 2.3.b: PUB 2.3.c: PRI 100% PUB

4.1

Enhancement of the Competitiveness of Local Enterprises

227,518,870

113,759,435

4.2

Public Administration for Enterprises

9,000,000

4,500,000

4.4

Integrated Development of Local Production Chains

10,000,000

5,000,000

4.5

Enhancement and Improvement of Sardinian Tourism Sector

110,456,000

55,228,000

319,068,000 28,778,000

159,534,000 14,389,000

3.13.a: PUB 3.13.b: PRI 4.1.a: PRI 4.1.b: PRI 4.1.c: PRI 4.1.d: PRI 4.1.e: PRI 4.1.f: PRI 4.1.g: PUB 4.2.a: PUB 4.2.b: PRI 4.2.c: PUB 4.2.d: PRI 4.4.a: PPP 4.b: PRI 4.5.a: PRI 4.5.b: PUB 4.5.c: PUB 4.5.d: PRI 4.5.e: PUB 100% PUB PUB and PRI

201,572,566

100,786,283

100% PUB

98,251,434

49,125,717

100% PUB

94,660,000

47,330,000

Safety Issues and Control Legality of Investments

12,000,000

6,000,000

Technical Assistance TOTAL

23,450,000 2,600,980,000

11,725,000 1,300,490,000

100% PUB 6.5.a: PUB 6.5.b: PUB 6.5.c: PUB 6.5.d: PRI 100% PUB

3.13

5.1 5.2 6.1 6.2 6.3 6.5 7.1 a

Total cost [D ]

Policies Aimed at Urban Areas Quality of Life within Cities: Improvement of Social Services Multimodal Transport Network Linking Sardinia to the Mainland Transport Systems within Sardinia’s Greater Urban Areas: Accessibility and Transport Management Information Society

PUB: public authorities; PRI: private sector; PPP: public–private partnerships.

The following paragraphs describe sources of information and procedures used to fill in the above additional fields. 3.1. Resident population and population density “POP2001” and “POP2007” are two numeric fields. For each municipality, they provide information on resident population as of 20015 and as of December 31, 20076 respectively. A third numeric field (“POPCH0107”) was then added to show population change

5 Source: 14th Population and Housing Census, published by the Italian Institute of Statistics and accessible on the internet at http://dawinci.istat.it/daWinci/jsp/MD/dawinciMD.jsp (accessed 19.07.10). 6 Source: the official website of the Regional Administration of Sardinia www.sardegnastatistiche.it (accessed July 19.07.10).

between 2001 and 2007; for each municipality, this value was derived by dividing the difference between the population in 2007 and 2001 by the 2001 population. A fourth and fifth numeric field (“DENS2001” and “DENS2007”) were also introduced. They give information about population density (derived by dividing the number of resident population by the land area) in each municipality in 2001 and 2007 respectively. Fig. 1 contains two choropleth maps in which Sardinian municipalities are classed according to the values of “POP2007” and “DENS2007.” 3.2. Coastal and inland municipalities “COAST” is a Boolean field and distinguishes coastal cities and towns from inland ones. Cells belonging to this field were filled

476

C. Zoppi, S. Lai / Land Use Policy 28 (2011) 472–485

Fig. 1. Resident population in 2007 (left) and population density in 2007 (residents per km2 , right).

by means of a spatial query (Biallo, 2002; Korte, 2001), or in other words, by means of a query based on a spatial relationship between two features. A shapefile containing the coastline was therefore introduced to perform the query, which consisted in identifying those municipalities whose boundaries meet the coastline (proximity operator). This made it possible to label as “coastal” 73 out of the 377 Sardinian municipalities, whose spatial distribution is shown in Fig. 2. 3.3. A categorization of municipalities The field “TYP CITY” categorizes Sardinian municipalities into five groups, as defined by the Sardinian 2007–2013 ROP-ERDF (pages 151 and 152). These groups are as follows: • • • •

Greater urban areas. Mid-dimensional urban areas. Mountain towns and cities. Municipalities included in a Local Work System (LWS) having a population density in 2001 greater than or equal to 50 inhabitants per km2 . • Municipalities included in a Local Work System (LWS) having a population density in 2001 smaller than 50 inhabitants per km2 . As for “greater urban areas,” the Sardinian 2007–2013 ROPERDF lists 23 municipalities as belonging to this group for their importance at the regional level. Only municipalities very close to the two main cities in the island (that is, Cagliari and Sassari) are included in this group, which comprises, in addition to the aforementioned cities, Alghero, Assemini, Cagliari, Capoterra, Castelsardo, Decimomannu, Elmas, Maracalagonis, Monserrato, Porto Torres, Pula, Quartu Sant’Elena, Quartucciu, Sarroch, Sassari, Selargius, Sennori, Sestu, Settimo San Pietro, Sinnai, Sorso, Stintino, and Villa San Pietro. Ten municipalities (Carbonia, Iglesias, Lanusei, Nuoro, Olbia, Oristano, Sanluri, Tempio Pausania, Tortolì, and

Villacidro) are listed as “mid-dimensional urban areas” in the Sardinian 2007–2013 ROP-ERDF. In contrast to cities belonging to the first group, mid-dimensional urban areas have a smaller importance, somewhat limited to the local level. The third group comprises mountain towns and cities, as defined by the DRG no. 49/16 of October 21, 2005. Following Sardinian Regional Law no. 12/2005, this DRG lists as mountain cities (a) those municipalities whose land area has an altitude of at least 400 m above the sea level for at least a half of its surface, and (b) those municipalities whose land area has an altitude of at least 400 m above the sea level for at least a 30 percent of its surface, provided that the difference between minimum and maximum height is at least equal to 600 m. Since seven (Iglesias, Lanusei, Maracalagonis, Pula, Sinnai, Tempio Pausania, Villacidro) out of the 120 municipalities classed as mountain cities by the DRG had already been categorized according to the ROP-ERDF either as greater urban areas or as mid-dimensional urban areas, only 113 municipalities were here considered as included in this third group. The 231 remaining municipalities were classed on the basis of the value of the population density (as of 2001) of the LWS they belonged to.7 In order to calculate this value, the shapefile containing municipal boundaries was used together with a twofield matrix linking each municipality with the LWS it belonged to.8 The dissolve procedure (Chrisman, 2002), also known as “spatial aggregation” (Biallo, 2002), was then used to aggregate all the polygons (in this case, those representing Sardinian municipalities) that shared the same value of a certain attribute (here, the LWS codes). By doing so, boundaries dividing two adjacent homogeneous polygons were dropped and a new shapefile, repre-

7 A spreadsheet listing Italian LWS’s is available on the Internet at the Italian Institute of Statistics’ (ISTAT) website: http://dwcis.istat.it/cis/docs/ sistemi/tav 12 sll.xls (accessed July 19.07.10). 8 The table can be downloaded from http://www.sardegnastatistiche.it/ documenti/12 166 20080314111709.xls (accessed July 19.07.10).

C. Zoppi, S. Lai / Land Use Policy 28 (2011) 472–485

477

Fig. 2. Coastal cities and towns.

senting the 45 Sardinian LWS, was obtained. Land area and resident population of each LWS were derived from municipal land areas and municipal resident populations by means of the summarize function9 in order to calculate, for each LWS, its population density. This made it possible to categorize our remaining 231 municipalities according to the value of this attribute. It is worth noting, though, that an exception was made for the Oristano LWS. In fact, this LWS (whose population density is greater than 50 inhabitants per km2 ), according to the ROP-ERDF is characterized by a significant level of disparity between its constituent cities, so that two groups can be identified. On the one hand lies the main city (Oristano) and its hinterland, consisting of four cities; on the other hand, 20 disadvantaged, small-sized towns and villages having an extremely low population density. These 20 municipalities were thus categorized as belonging to the fifth group (“municipalities included in a Local Work System (LWS) having a  population density in 2001 smaller than 50 inhabitants per km2 ), in spite of the actual population density of their LWS, to take into account their status as disadvantaged areas. The choropleth map in Fig. 3 shows the output of this classification process.

9 Given a table where a field (in which multiple occurrences of a certain value are allowed) has been selected as the key field, the summarize function returns an output summary table consisting of as many rows as the unique values of the selected field in the original table are. Various statistical fields can be added to the summary table (Environmental Systems Research Institute, 1999; Hutchinson, 2004).

Fig. 3. Sardinian municipalities: advantaged and disadvantaged areas.

3.4. Financial data Monitoring data related to projects co-financed through public funds (be they from the European Union, from the Italian central government or from the Sardinian regional government) have been made available to the general public via the official website of the Regional Administration of Sardinia.10 Various search functions have also been developed; one of these functions allows users to perform a specific query based on the measure of the Sardinian ROP 2000–2006 through which projects were funded. This required a preliminary analysis of the measures of the Sardinian ROP 2000–2006, in order to identify those that had been funded by the ERDF (Table 1).11

10 A dedicated section of the site, titled “Progetti in corso” (“Ongoing projects”) is available at http://progetti.regione.sardegna.it/argomenti/progetti/ (accessed July 19.07.10). 11 In Table 1 all of the measures in the Sardinian ROP 2000-1006 funded through the ERDF are listed; however, for our research, we did not take measure 7.1 (“Technical Assistance”) into account. The reason for this exclusion is twofold: on the one hand, technical assistance was a cross-cutting measure, which supported all axes and measures in the ROP (including those funded through other sources of funding); on the other hand, it was mainly implemented at the regional level, thus it evenly benefited Sardinian cities and towns.

478

C. Zoppi, S. Lai / Land Use Policy 28 (2011) 472–485

For each of the 20 measures co-funded by the ERDF, a database was downloaded from the website. This database contained all of the projects funded through a certain measure; projects in the database were uniquely identified by means of a code and a title, and described by means of some attributes, such as the place where the project is (or was) carried out, the amount of the projected investment, and the share of the projected investment already spent by August 31, 2008. These 20 databases were subsequently aggregated, which led to the creation of a unique database containing 9128 rows, one for each project co-funded by the ERDF, irrespective of the corresponding measure of the ROP 2000–2006. 350 out of these 9128 projects were deleted because they were not related to a single municipality; in fact, 294 were carried out at the regional level and affected therefore the whole Sardinian territory; 54 were carried out at the provincial level, while two took place outside Sardinia. As a result, 8778 projects could be directly connected to a municipality. However, 126 out of these 8778 were deleted because the field “projected investment” had not been correctly populated (for 65 records the field had not been filled; for the remaining 61 records, the value was zero). The number of projects that could be used for the purpose of this research was therefore equal to 8652. By means of the summarize function, and by choosing as the key field the one containing the codes that uniquely identify each municipality, a summary table consisting of 364 records was obtained. This means that 13 out of the 377 Sardinian municipalities had been left out, either because no project cofunded by the ERDF had been carried out in those cities or towns, or because the field “projected investment” had not been correctly populated. For each of the 364 municipalities appearing in the summary table, the total projected investment (“PROJ INV”) and the total actual expenditure as of August 2008 (“EXPEND”) were chosen as descriptive statistics. Finally, for each municipality, expenditure efficiency (“EFFIC”) was calculated as total projected investment divided by total actual expenditure (Fig. 4). In the following section, data contained in the spatial database built by means of the GIS-based analysis presented in this section are fed into the MNL model, so as to examine the impact of the attributes here identified and quantified on expenditure efficiency of Sardinian cities and towns by means. 4. Some interpretive observations on the expenditure efficiency based on an MNL model application This section is organized as follows. In the first paragraph, the MNL model methodology is presented in the context of the case study discussed in this essay. Secondly, the results concerning Sardinian Region’s expenditure efficiency in the implementation of the 2000–2006 ROP-ERDF Sardinian Region are discussed; there results stem from the application of the MNL model methodology. 4.1. Methodology MNL models describe how people choose among a discrete set of mutually exclusive alternatives. McFadden’s work (1978, 1980) on generalized extreme value formulation, which generalized the work of Williams (1977), provides a rigorous foundation for consumer choice modeling derived from economic theory. Although the original formulation of the random utility maximization as a behavioral model followed the economists’ theory of consumer behavior, it also included features of the taste template that were

heterogeneous across individuals and unknown to the analyst, as well as unobserved aspects of experience and of information on the attributes of alternatives, interpreted as random factors (McFadden, 1978, 1980, 2000). This led to the paradigm for generating discrete-choice models commonly reported in textbooks (Ben-Akiva and Lerman, 1985; Ortúzar and Willumsen, 2001; Train, 2009), that the random part of the individual utility reflects the modellers’ lack of complete information about all the elements considered by the individual making a choice and the observed deviations of individual behavior from perfect rationality (Tversky, 1972). The key assumption of the MNL models is that the errors are independent of each other. This independence means that the unobserved portion of utility for one alternative is unrelated to the unobserved portion of utility for another alternative. It is a fairly restrictive assumption, and the development of other models has arisen largely for the purpose of avoiding this assumption and allowing for correlated errors. It is important to realize that the independence assumption is not as restrictive as it might at first seem, and in fact can be interpreted as a natural outcome of a well-specified model.12 MNL models are generally used to study phenomena characterized by nominal observations, that is, observations represented by categories of outcomes defined by means of names. These names do not represent any order. The assumption of the MNL models is that these phenomena are correlated to other phenomena, represented by numerical and nominal variables, through a logistic probability function. This function makes it possible to characterize these correlations. MNL models are used in different and mostly-heterogeneous scientific fields, since they are extremely effective, theoretically and empirically, in order to analyze several issues concerning the interpretation of human behavior.13 MNL models are widely used to study subjective choices between multiple alternatives. This is true, for example, when referring to the choice between different destinations for recreational activities. Bockstael et al. (1987) use an MNL model to characterize the choice between seawater and freshwater beaches, while Bockstael et al. (1991), when describing the approach proposed by McFadden (1974a, 1978), put in evidence that his methodology requires particular attention if the choice implies nested steps, such as “if . . . then . . . else . . .,” which follow each other: choice of a recreational activity (e.g. fishing), choice of the activity type (e.g. underwater fishing), choice of the most suitable, choice of the type of fish, etc. Consistent with this framework is McFadden’s theoretical and practical research, which was awarded the Nobel prize in 2000 (McFadden won the prize with Heckman). McFadden studies the issue of discrete choice under different points of view. Among these are the demand for urban transport services (McFadden, 1974b), the choice between different transport modes (McFadden and Train, 1978), the demand for local phone services (McFadden et al., 1987), the decision-making processes of the public administration (McFadden, 1976). Other important applications of MNL models are referred to the analysis of wage mobility in Europe (Pavlopoulos et al., 2010), and to the assessment of the incidence of specific external factors on the ineffectiveness of particular clinical treatments (Ambrogi et al., 2009). In this essay, an MNL model is used to analyze the relation between a discrete variable, the expenditure efficiency of the

12 13

This introductory discussion largely draws on Cherchi (2009). For the applications of MNL models in several fields see Bhat (2007).

C. Zoppi, S. Lai / Land Use Policy 28 (2011) 472–485

479

Fig. 4. Sardinian municipalities: projected investments referring to projects co-funded by the ERDF (left) and expenditure efficiency (right).

2000–2006 ROP-ERDF, and other variables that are likely to be correlated to the efficiency variable. The MNL model is based on Greene (1993, pp. 666–672), which draws on the Nerlove and Press’s (1973) approach. The model considers a set of events J, J = {0, 1, . . ., N}, with probability of event Yi = j given by Eqs. (1) and (2)14 : e

Prob(Yi = j) =

1+

Prob(Yi = 0) =

ˇ xi j

j

e k=1

ˇ xi

,

1

1+

j

k=1

e

(1)

j

ˇ xi

,

(2)

k

where j ∈ {1, . . ., N}, ˇj is a vector of coefficients referred to the event j, and xi is a vector of characteristics of the territorial context i, where the event j occurs, i ∈ {1, . . ., M}. Coefficients ˇj s are estimated by solving the maximization problem of the following log-likelihood function, ln L: ln L =

M M  

dij ln Prob(Yi = j)

(3)

i=1 j=0

where dij = 1 if in the context i the event j occurs, and dij = 0 otherwise), in the coefficients ˇj s. These coefficients will appear in (3) through the expressions (1) and (2) of Prob(Yi = ·).

14 If we define ˇj * = ˇj + q for any nonzero vector q, the identical set of probabilities result, as the terms involving q all drop out. A convenient normalization that solves the problem is to assume that vector ˇ0 = 0. The probability for Y = 0 is therefore given by (2) (Greene, 1993, p. 666).

The derivatives of (3) with respect to the coefficients ˇj s have the following form:

M ∂ln L = [dij − Prob(Yi = j)]xi ∂ˇj i=1

(4)

The values of the vectors of coefficients ˇj s which maximize (3) are the solution of the system which comes from equalizing to zero the derivatives expressed by (4). The values of the vectors of coefficients ˇj s make it possible to calculate the marginal effects of a change of the vector of characteristics xi on the probability that the event j occurs in the context i, ∂Prob(Yi = j)/∂ xi , as follows: ∂Prob(Yi = j) = [Prob(Yi = j) ∂xi



ˇj −

J k=1



[Prob(Yi = k)]ˇk

. (5)

The estimate of the model makes it possible to calculate the marginal effects of (5), e.g. with reference to the average values of the xi s, and the probabilities of the events js. Moreover, the model makes it possible to estimate the standard errors of the estimates of the ˇj s and of the marginal effects of (5). 4.2. The expenditure efficiency analysis The analysis of the expenditure efficiency of the Sardinian regional administration for the implementation of the investment policies of the 2000–2006 ROP-ERDF is based on the model described in Section 4.1. Characteristics xi and their measures are defined, and the results of the model application are reported with reference to: (i) the estimates of coefficients ˇj s, (ii) the marginal

480

C. Zoppi, S. Lai / Land Use Policy 28 (2011) 472–485

Table 2 Definition of variables and descriptive statistics. Variable

Definition

Mean

S.D.

EFFIC

Discrete variable (variable Yi in formulas (1)–(5) of Section 4) – expenditure efficiency calculated as the fraction of the 2000–2006 projected investment spent by August 31, 2008 – 0 if it is less than the 20th percentile; 1 if it is between the 20th and the 40th percentiles; 2 if it is between the 40th and the 60th percentiles; 3 if it is between the 60th and the 80th percentiles; 4 if it is greater than the 80th percentile – the 20th, 40th, 60th and 80th percentiles are the following: 39.72; 57.25; 71.99; 86.51 City resident population (number of residents) City residential density (residents per km2 ) Dummy – greater urban area, so-defined by the 2007–2013 ROP-ERDF (p. 152) Dummy – medium-sized urban area, so-defined by the 2007–2013 ROP-ERDF (Commission of the European Communities, 2008) Dummy – city included in a Local work system (LWS) with a residential density in 2001 greater than or equal to 50 residents/km2 Dummy – city included in an LWS with a residential density in 2001 greater than or equal to 50 residents/km2 Dummy – the city boundary does not coincide at all with the seashore Projected investment of the city from the 2000–2006 ROP-ERDF (D )

62.7567

25.0856

4,565.8819 77.5029 0.0632

12,440.2175 212.4531 0.2436

0.0275

0.1637

0.2198

0.4147

0.3791

0.4858

0.7995

0.4010

10,480,952.1974

32,530,086.1987

POP07 DENS07 GRCITY MIDCITY LWSHD

LWSLD COAST AMOUNT

Variables POP07, DENS07, GRCITY, MIDCITY, LWSHD, LWSLD, COAST and AMOUNT are the components of vectors xi s in formulas (1)–(5).

effects from (5), and (iii) the probabilities of the events contained in the set J. The territorial contexts, indicated by letter i in Section 4.1, i ∈ {1, . . ., M}, are the 364 Sardinian cities whose policies were funded by the 2000–2006 ROP-ERDF. These are the vast majority of the Sardinian cities, since only thirteen cities were excluded from the program. Characteristics xi , their measures and descriptive statistics are reported in Table 2. The discrete variable which describes the expenditure efficiency, indicated by Yi in Section 4.1 and labeled EFFIC in Table 2, takes five values (0, 1, 2, 3, 4) identified by the five intervals of the 20th, 40th, 60th and 80th percentiles. The choice of the explanatory variables which characterize the 364 Sardinian cities is motivated as follows. The resident population (POP07) is connected to the social, economic and political relevance of the city, while the residential density (DENS07) puts in evidence, at least to a certain extent, the relation between residential layout and a city’s territory: it is evident that, everything else being the same, the greater the residential density, the greater the consumption of the urban territory, and the lower the availability of areas for future urban residential expansions. Moreover, a more consumed urban territory implies a greater attention to the protection of natural resources and to urban renewal policies. A typological taxonomy of the Sardinian cities (variables GRCITY, MIDCITY, LWSHD, LWSLD) is based on the urban development concept and the policies for the disadvantaged zones of the 2007–2013 ROP-ERDF, which are referred to the geographic concentration of the investments (see the Introduction). These variables put in evidence, with reference to the policies implemented between 2000 and 2006 if, and by how much, the efficiency of the expenditure implemented in the disadvantaged zones is different from the efficiency of the expenditure for the urban areas, and so, if, and by how much, investment policies for territorial cohesion, based on the principle of geographic concentration, have been properly implemented. The variable that indicates if a city is coastal or non-coastal aims at assessing if the expenditure efficiency is influenced by this char-

acteristic, which is relevant because, since November 2004, when Regional Law no. 8 was approved, Sardinian coastal cities have had to make their planning policies consistent with very strict planning rules, especially for areas included in the coastal zone (Zoppi, 2008).15 The variable which represents the projected investment of the city from the 2000–2006 ROP-ERDF characterizes the relative importance that the ROP recognizes to a particular city, and the level of trust of the regional administration in the efficiency of the city in implementing its expenditure policies. The procedure described in the previous paragraph is used to estimate the coefficients ˇj s and marginal effects ∂Prob(Yi = j)/∂ xi on the probability of the events js at the mean values of the xi s. The estimates of the marginal effects on the probabilities of the events js at the mean values of the xi s and the cumulative probabilities at the mean values of the xi s are reported in Table 3. The estimates of the coefficients ˇj s that solve the maximization problem expressed by (3) are reported in Table 4 (Appendix 1). Two tests concerning the goodness-of-fit of the MNL model have been implemented. The result of a standard log-likelihood test gives no evidence of lack of fit (see the last row of Table 4, Appendix 1), since the chi-square statistics corresponding to the log-likelihood ratio is close to 15 percent, if we test the null hypothesis that the distributions of the observed and expected occurrences of Yi are not different from each other. Moreover, a Hosmer and Lemeshow (1989) test is implemented (see Table 5, Appendix 2). This compares the expected versus observed occurrences of events Yi = 0 (very low efficiency), Yi = 1 (low efficiency), Yi = 2 (medium efficiency), Yi = 3 (high efficiency), Yi = 4 (very high efficiency). Data used to implement the test are reported in Table 5 (see Appendix 2) for the event Yi = 1. The test

15 This law (named “The saving-coast law”) identifies the coastal areas as those included in a 2-km belt from the coastal line. The definition of the coastal zone changed after the Regional Landscape Plan (RLP) approval (through the Deliberation of the Sardinian Regional Government no. 36/7 of September 5, 2006), which established a new and generally wider limit for the coastal zone.

C. Zoppi, S. Lai / Land Use Policy 28 (2011) 472–485

481

Table 3 Marginal effects on the probabilities of the five events which define expenditure efficiency. Variable

Marginal effect

t-Statistic

Hypothesis test: marginal effect = 0

Marginal effect on probability of Yi = 0 (very low efficiency), ∂Prob(Yi = 0)/∂xi , Prob(Yi = 0) = 0.245 POP07 −9.6763E−06 −3.990 DENS07 5.8289E−05 0.282 GRCITY 2.2000E−01 0.957 MIDCITY −7.6243E−03 −0.009 LWSHD 5.0301E−02 2.610 LWSLD −9.9512E−02 −5.417 COAST −8.5391E−02 −2.545 AMOUNT 2.8552E−09 0.283 Marginal effect on probability of Yi = 1 (low efficiency), ∂Prob(Yi = 1)/∂xi , Prob(Yi = 1) = 0.247 POP07 1.3677E−05 2.370 DENS07 1.6634E−04 0.752 GRCITY −8.2628E−02 −0.312 MIDCITY 2.5281E−01 0.532 LWSHD 2.4307E−03 0.088 LWSLD −4.2738E−02 −4.070 COAST −6.0760E−02 −2.012 AMOUNT −5.8521E−09 −0.679 Marginal effect on probability of Yi = 2 (medium efficiency), ∂Prob(Yi = 2)/∂xi , Prob(Yi = 2) = 0.255 POP07 1.2930E−06 0.308 DENS07 1.0083E−04 0.449 GRCITY −1.3537E−01 −0.5 MIDCITY −1.3144E−01 −0.261 LWSHD 1.5503E−02 0.606 LWSLD 1.8582E−02 9.567 COAST 1.0740E−01 15.996 AMOUNT 2.1256E−09 0.202 Marginal effect on probability of Yi = 3 (high efficiency), ∂Prob(Yi = 3)/∂xi , Prob(Yi = 3) = 0.244 POP07 −4.2481E−06 −1.347 DENS07 −2.7179E−04 −1.670 GRCITY 9.0627E−02 0.362 MIDCITY −6.9910E−02 −0.144 LWSHD −7.5626E−02 −2.007 LWSLD 1.2256E−01 8.906 COAST 3.2878E−02 2.001 AMOUNT 3.4977E−09 0.345 Marginal effect on probability of Yi = 4 (very high efficiency), ∂Prob(Yi = 4)/∂xi , Prob(Yi = 4) = 0.010 POP07 −1.0455E−06 −0.168 DENS07 −5.3672E−05 −0.058 GRCITY −9.2629E−02 −0.045 MIDCITY −4.3830E−02 −0.005 LWSHD 7.3916E−03 0.102 LWSLD 1.1119E−03 0.098 COAST 5.8678E−03 0.047 AMOUNT −2.6263E−09 −0.104

compares the number of the observed and expected occurrences of Yi = A by partitioning the observations into 10 equal-sized groups on the basis of increasing predicted probabilities. With reference to this partition, the Hosmer and Lemeshow (HL) statistic is calculated, as follows: HL(Yj = A) =

10  j=1

(Oj A − Ej A )

2

Ej A (1 − (Ej A /nj A ))

(6)

where A takes five values (0, 1, 2, 3, 4), Oj A and Ej A are the number of observed and expected occurrences of event A in the jth group, and nj A is the number of observations in the jth group of partition concerning Yi = A, j ∈ {1, . . ., 10}. The Ej A s are calculated through the expressions (1) and (2) of Prob(Yi = ·). The Hosmer and Lemeshow test consists of testing the HL statistic by a chi-square test with 8 degrees of freedom. Since the values of the HL statistic, reported in Table 5 (see Appendix 2), are always greater than 10 percent, the Hosmer and Lemeshow test shows no evidence of lack of fit, if we test the null hypothesis that the distributions of the Oj A s and Ej A s are not different from each other. In the following paragraphs, the results concerning the effects of the explanatory variables on the five events which describe

0.0004 0.7798 0.3457 0.9929 0.0137 0.0000 0.0319 0.7790 0.0240 0.4575 0.7571 0.5984 0.9304 0.0003 0.0527 0.5020 0.7601 0.6565 0.6205 0.7958 0.5488 0.0000 0.0000 0.8412 0.1874 0.1047 0.7197 0.8864 0.0533 0.0000 0.0539 0.7324 0.8676 0.9541 0.9644 0.9960 0.9194 0.9225 0.9628 0.9178

expenditure efficiency (very low, low, medium, high, very high) are discussed. It has to be underlined that the “very high” event is highly improbable at the mean values of the explanatory variables of the Sardinian cities (the probability is about 1 percent). For this reason, we consider only the other four events in the discussion below. 4.3. Resident population The influence of the resident population is ambiguous, since the marginal effect is negative and significant16 on the probability of the event “0” (a very low expenditure efficiency), while it is positive and significant on the probability of the event “1” (a low expenditure efficiency). The marginal effect on the probability of the event “2” (a medium expenditure efficiency) is positive and non-significant, while it is negative and non-significant on the probability of the event “3” (it would be significant at a 20 percent hypothesis test). In quantitative terms, the marginal effect on the probability of the event “1” is greater than the marginal effect

16 If not otherwise specified, we define “significant” a value of a marginal effect which is significant at a 5 percent hypothesis test.

482

C. Zoppi, S. Lai / Land Use Policy 28 (2011) 472–485

on the probability of event “0”; the same for the event “3” with respect to the event “2.” Hence, albeit ambiguous, the marginal effect of the variable resident population seems to be more negative than positive on the probability of the highest values of the expenditure efficiency, and vice-versa on the probability of the lowest values. 4.4. Population density The marginal effect of the population density is positive and non-significant on the probability of events “0,” “1” and “2,” while it is more significant (at a 8 percent hypothesis test) for the event “3.” This indicates that this variable is certainly more influential on medium-high efficiency, and it seems to be negatively correlated to the probability of this event.

4.9. Non-coastal cities As for the latter case, marginal effects are always significant, and negative on the events “0” and “1,” and positive otherwise. This means that non-coastal cities are more virtuous than coastal cities. 4.10. Projected investment of the city from the 2000–2006 ROP-ERDF Marginal effects of the Projected investment of the city from the 2000–2006 ROP-ERDF area always non-significant. Their signs are positive in case of events “0,” “2” and “3,” and negative in the other case. Hence, there is no indication of an influence of this variable on expenditure efficiency.

4.5. The city belongs to a greater urban area The marginal effect of the variable which puts in evidence that a city belongs to a greater urban area as classified by the ROP-ERDF 2007–2013 (p. 152) instead of a disadvantaged zone, defined by the Article 10 of Regulation no. 1080/2006/EC with reference to the Article 52 Regulation no. 1083/2006/EC – which, in the Sardinian case, are identified with mountain areas, and zones with low (less than 50 inhabitants per km2 ) and very low (less than 8) demographic density – is always non-significant, positive on the probability of events “0” and “3,” and negative otherwise. Since the values of the marginal effects of this variable are always non-significant, it can be assumed that there is no evidence of an influence of this variable on the expenditure efficiency. 4.6. The city belongs to a medium-sized urban area Considerations related to the results obtained for the variable which puts in evidence that a city belongs to a medium-sized urban area are almost the same as in the case of variable “GRCITY.” Thus, for that variable there is no evidence of an influence on the expenditure efficiency as well. 4.7. The city is included in an LWS with a residential density in 2001 greater than or equal to 50 residents/km2 The marginal effect of the dichotomous variable which puts in evidence that a city belongs to an LWS with a high residential density with respect to being included in a disadvantaged zone, is positive and significant on the event “0,” and negative and significant on the event “3.” In the other two cases, the marginal effect is positive and non-significant, with very low values of the t-statistics of the hypothesis tests. Therefore, the results indicate that cities included in high-density LWSs are comparatively less virtuous in terms of expenditure efficiency than cities belonging to disadvantaged zones. 4.8. The city is included in an LWS with a residential density in 2001 lower than 50 residents/km2 The marginal effect of the dichotomous variable which puts in evidence that a city belongs to an LWS with a low residential density with respect to being included in a disadvantaged zone, is negative and significant on events “0” and “1.” In the other two cases, the marginal effect is positive and significant. This implies that cities included in low-density LWSs are comparatively more virtuous in terms of expenditure efficiency than cities belonging to disadvantaged zones.

5. Discussion and conclusion The results put in evidence that, in general, expenditure efficiency is a real problem concerning the 2007–2013 ROP-ERDF if one considers the implementation of the 2000–2006 ROP-ERDF. There are three major implications of the MNL model results for a better understanding of territorial phenomena and a significant improvement in the effectiveness of Sardinian ROP-based policies. First, it can be noticed that expenditure is highly efficient if the resident population and its density are low, the city is noncoastal and belongs to a disadvantaged LWS. This indicates that cities having more residents and a higher population density – at least for high levels of expenditure efficiency – and included in a less disadvantaged LWS, are less efficient. This observation strongly suggests that the implementation of the principle of geographic concentration of investment as it is applied by the 2007–2013 ROP-ERDF (Article 37(3) of Regulation (EC) no. 1083/2006) should be revised. The ROP establishes that investment should be evenly shared between large- and medium-sized urban areas, and disadvantaged zones, as defined above. The results of the MNL model show that only less-populated and low-density cities are reliable in terms of expenditure efficiency, so only these cities should be targeted for geographically concentrated investment. From this point of view, there is a strong case for concentrating investment in small-sized towns and for decreasing the investment share in large- and medium-sized urban areas, in a future revised version of the 2007–2013 ROP-ERDF. Second, the MNL model’s results show that cities having at least a part of their administrative boundary coincident with the seashore are comparatively less efficient. This can be easily explained with reference to the planning implementation code (PIC) of the RLP, which establishes very strict and conservative rules for the coastal cities: this has very possibly exercised a negative influence on the administrative capability of these cities, and on their efficiency in spending. Moreover, the depressing impact of the restrictive rules of the PIC of the RLP on public investment could be possibly connected to other negative impacts. Coastal cities could suffer from the decline of building expansion rights, since they could not rely any longer on the financial resources for public services and infrastructure that would come from the impact fees paid by the developers. Another problem for the budget of the city would come from the decrease in payments of the communal tax for real estate which includes land property, since the value of land would dramatically drop without development rights. Since in many of the actual tourist coastal zones it would not be possible to build anymore, a crisis of the local construction industry would probably occur. This industry is the most important in terms of

C. Zoppi, S. Lai / Land Use Policy 28 (2011) 472–485

income and employment for the local economy, which is characterized by a high unemployment rate. Its crisis would worsen an already difficult economic and social situation (Regione Autonoma della Sardegna, 2006a,b). Third, the results show that a more virtuous behavior is not connected with the fact that a city belongs either to greater or medium-sized urban areas, which will benefit from a significant part of the 2007–2013 ROP-ERDF projected investment, following what established by Article 8 of Regulation no. 1080/2006/EC. This finding indicates that the urban dimension of the EU cohesion policy of the Sardinian ROP should not be based on city size, as it is in the actual version of the program. Investment in the urban contexts should aim to address and possibly solve problems that may arise in specific local contexts, which are located inside the cities. These problems are site-specific, neighborhood-specific and community-specific, and can hardly be identified and correctly analyzed at the whole-city level. This implies a massive involvement of local communities in the ROP implementation as it is discussed in the “Agenda for a reformed cohesion policy” (Barca, 2008), and by the Commission of the European Communities (2008), which states that “local involvement is essential to reach a high degree of acceptance and visibility on the ground and concerns not only integrated operations, but also sector-oriented activities in cities. The programming documents for 2007–2013 generally show few signs of direct local involvement in the design and implementation of ERDF Operational Programmes” (p. 5). Policy recommendations from the findings of this study can be discussed in the context of the analysis of the urban dimension of the ERDF operational programs proposed by the Commission of the European Communities (2008). These recommendations refer to three issues. First, as we mentioned above, small-sized urban areas are comparatively more efficient than large- and medium-sized. So, it would be recommended that the ERDF policy should concentrate investment in these areas, which belong to weak local work systems. This implies that a “Lisbon-oriented” investment should flow to small-sized urban areas (Commission of the European Communities, 2008, p. 30). A “Lisbon-oriented” investment is one which emphasizes the role of entrepreneurship, innovation and support to small-sized enterprises, in order to implement the Strategy of Lisbon of the European Union (European Commission, 2010). In other words, findings from this study suggest that a profound revision of the ERDF policy should take place, so as not to concentrate funds on large and medium-sized urban areas anymore. Concentration of investment in small-sized deprived areas may eventually catalyze local economic and social development of the other urban areas, as a by-product of this more selective concentration policy. A second important policy recommendation from the findings of this study concerns the policy of network support for a balanced, polycentric development (Commission of the European Communities, 2008, pp. 31–32).17 Network support consists of operations to improve networking between large-, medium-, and small-sized urban areas, and rural centers as well, in order to develop cooperation between cities and their hinterlands. Networking implies improved accessibility conditions in terms of transportation systems, water resources, energy, ICT, etc. Improved networking would imply that a widespread cooperative approach between urban areas, characterized by a different size, local work

17 This question is also discussed in the following documents: (i) “ESDP. European Spatial Development Perspective” (European Commission, Committee on Spatial Development, 1999); (ii) “Towards a Thematic Strategy on Urban Environment” (CEC, 2004a).

483

system and expenditure efficiency, could encourage a more balanced and widespread urban dimension at the regional level, much to the benefit of all of the cities. Good practices in terms of expenditure ability, which characterize small-sized urban areas, could be shared, and small-sized urban areas could benefit from a close relationship with large- and medium-sized ones in terms of a more balanced and polycentric development, based on concentration of public investment in city networks rather than in large- and medium-sized cities. This would also improve the depressed local work systems of depressed small-sized Sardinian urban areas. A third recommendation stems from urban renewal and environmental rehabilitation. These are policies the European Union considers fundamental to foster the urban dimension of the ERDF-funded regional operational programs (Commission of the European Communities, 2008, pp. 30–31). Urban historical heritage preservation and renewal and natural environment protection are policies that concern all of the cities, large- and medium-sized, and small-sized as well, and the ERDF strongly relies on investment for the implementation of these policies. The findings of this study indicate that there is no need for investment concentration on large- and medium-sized cities, since good practices for historical centers renewal and natural environment rehabilitation could spread if networking between differently sized urban areas were fostered. In other words, ERDF investment should concentrate in fostering cooperation between cities of various size, instead of flowing onto large- and medium-sized cities, whose investment ability is poor. The implementation timing and development of cooperative approaches to policy management with the city administrations, in order to make the ROPs more efficient in terms of expenditure, are also fundamental. Moreover, a particular attention is to be devoted to administrative procedures, which involve both the regional administration and the cities, as they generated slowdowns and inefficiencies during the implementation of the 2000–2006 ROP. This paper has employed an MNL model to analyze the expenditure efficiency of Sardinian cities with respect to the 2000–2006 ROP-ERDF, building upon the results of a GIS-based analysis. In doing so, it demonstrates how a spatial analysis approach based on a GIS can be utilized to figure out the geography of territorial phenomena and to provide the MNL model with its database, thereby improving upon the objectivity and accuracy of the implemented MNL model. By doing so, the application of this method allows for an integration of the GIS and MNL approaches, which can be used by city planners in the development of policy-making processes concerning city residential areas. In this respect, the paper makes an important methodological contribution. By applying the method developed in this paper, planners can improve significantly their understanding of spatial phenomena, and the effectiveness of policy making and implementation. The technical and administrative procedures implemented by the Italian Regions to develop their ROPs, especially by the Regions of the Objective 1 of the 2000–2006 Cohesion Policy of the European Union, are quite similar to each other, since they are based on the same regulations of the European Union. For this reason, the results obtained by the GIS-based MNL model could be an important reference point to compare the expenditure efficiency of the Italian Regions of the Objective 1. In other words, an important feature of the methodology developed in this paper is that it is easily exportable, and, as a consequence, it allows for comparisons of different spatial configuration and policies. The optimal choice of the attributes to be included in the MNL model includes as many variables as necessary to describe the expenditure efficiency satisfactorily. Of course, this choice is heavily influenced by available information. The analysis here

484

C. Zoppi, S. Lai / Land Use Policy 28 (2011) 472–485

Table 4 Estimates of coefficients ˇj s that solve the maximization problem expressed by (3). Variable

ˇ coefficient

t-Statistics

Hypothesis test: ˇ = 0

Coefficients ˇj s for the discrete dependent variable Yi = 1 (low efficiency) POP07 9.4988E−05 2.082 DENS07 4.3585E−04 0.449 GRCITY −1.2343E+00 −1.465 MIDCITY 1.0558E+00 0.928 LWSHD −1.9579E−01 −0.428 LWSLD 2.3361E−01 0.530 COAST 1.0284E−01 0.245 AMOUNT −3.5390E−08 −1.888 Coefficients ˇj s for the discrete dependent variable Yi = 2 (medium efficiency) POP07 4.4631E−05 1.048 DENS07 1.5732E−04 0.131 GRCITY −1.4305E+00 −1.509 MIDCITY −4.8456E−01 −0.388 LWSHD −1.4481E−01 −0.315 LWSLD 4.7973E−01 1.107 COAST 7.7050E−01 1.695 AMOUNT −3.3325E−09 −0.233 Coefficients ˇj s for the discrete dependent variable Yi = 3 (high efficiency) POP07 2.2143E−05 0.524 DENS07 −1.3525E−03 −0.628 GRCITY −5.2787E−01 −0.623 MIDCITY −2.5543E−01 −0.202 LWSHD −5.1567E−01 −0.987 LWSLD 9.0925E−01 2.117 COAST 4.8388E−01 1.116 AMOUNT 2.6665E−09 0.194 Coefficients ˇj s for the discrete dependent variable Yi = 4 (very high efficiency) POP07 −6.6582E−05 −0.442 DENS07 −5.6874E−03 −0.966 GRCITY −1.0304E+01 −0.048 MIDCITY −4.4187E+00 −0.021 LWSHD 5.4480E−01 1.070 LWSLD 5.1971E−01 1.159 COAST 9.4482E−01 1.315 AMOUNT −2.7831E−07 −3.253 Coefficients ˇj s for the discrete dependent variable Yi = 0 (very low efficiency) are set to 0

0.0454 0.6565 0.1527 0.3604 0.6715 0.5998 0.8080 0.0681 0.3025 0.8966 0.1411 0.7006 0.7548 0.2765 0.0998 0.8172 0.6039 0.5345 0.5377 0.8412 0.3310 0.0421 0.2727 0.8474 0.6615 0.3413 0.9620 0.9834 0.2926 0.2550 0.1979 0.0027

Log-likelihood test. Log-likelihood ratio = 49.3083 − Prob. > chi-square = 0.14854.

implemented is based on a set of variables representing the best choice given the information available, rather than the optimal choice. These variables should be considered a subset of the optimal variable choice. Nevertheless, they give us an interesting picture of the phenomenon. Regarding this point, it must be stated that there are a number of variables that should have been included in the MNL model and were not included since no information is available. One is the household income, which could be very important in determining the income effect on the expenditure efficiency. Moreover, data on capacity of the system of public infrastructure and services would be very helpful. Acknowledgments I am enormously grateful to Romeo Danielis (University of Trieste, Italy) for the time he spent on studying the manuscript and his valuable comments on improving it. I thank Elisabetta Cherchi (Technical University of Denmark, Lyngby, Denmark) for her precious suggestions during the revision of this article. Appendix 1. See Table 4. Appendix 2. See Table 5.

Table 5 Hosmer and Lemeshow test for the goodness-of-fit of the estimated MNL model. Computational detail of the Hosmer and Lemeshow statistic HL(Yi = A) in case A = 1 (low efficiency). HL(Yi = A) is specified in (6) Group

Observed

Expected

1 2 3 4 5 6 7 8 9 10 Total

3 7 4 7 4 7 6 10 11 14 73

3.28 4.42 4.75 5.05 5.47 5.80 6.12 6.59 7.41 12.33 61.21

Chi-square test, 8 degrees of freedom

Prob. > chi-square

HL(Yi = 1) = 8.2027

0.41393

Chi-square test, 8 degrees of freedom, in case A = 0, 2, 3, 4

Prob. > chi-square

HL(Yi = 0) = 12.0061 HL(Yi = 2) = 12.5361 HL(Yi = 3) = 10.5333 HL(Yi = 4) = 12.8096

0.15093 0.12884 0.22953 0.11857

C. Zoppi, S. Lai / Land Use Policy 28 (2011) 472–485

References Ambrogi, F., Biganzoli, E., Boracchi, P., 2009. Estimating crude cumulative incidences through multinomial logit regression on discrete cause-specific hazards. Computational Statistics & Data Analysis 53, 2767–2779. Barca, F., 2008. An Agenda For A Reformed Cohesion Policy. A Place-based Approach to Meeting European Union Challenges and Expectations, Independent Report prepared at the request of Danuta Hübner, Commissioner for Regional Policy, available on the Internet at: http://ec.europa.eu/regional policy/policy/future/pdf/report barca v0306.pdf (accessed 19.07.10). Ben-Akiva, M., Lerman, S., 1985. Discrete Choice Analysis: Theory and Application to Travel Demand (Transportation Studies). The MIT Press, Cambridge, MA, United States. Bhat, C., 2007. Econometric choice formulations: alternative model structures, estimation techniques, and emerging directions. In: Axhausen, K.W. (Ed.), Moving Through Nets: The Physical and Social Dimensions of Travel – Selected Papers from the 10th International Conference on Travel Behaviour Research. Elsevier, Oxford, United Kingdom, pp. 45–80. Biallo, G., 2002. Introduzione ai Sistemi Informativi Geografici (An Introduction to Geographic Information System). MondoGIS, Rome, Italy. Bockstael, N.E., Strand, I.E., Hanemann, W.M., 1987. Time and the recreational demand model. American Journal of Agricultural Economics 69, 293–302. Bockstael, N.E., McConnell, K.E., Strand, I.E., 1991. Recreation. In: Braden, J.B., Kolstad, C.D. (Eds.), Measuring the Demand for Environmental Quality. North Holland, Amsterdam, The Netherlands, pp. 227–270. CEC (Commission of the European Communities), 2004. A New Partnership for Cohesion: Convergence Competitiveness Cooperation. Office for Official Publications of the European Communities, Luxembourg, Third Report on Economic and Social Cohesion. CEC (Commission of the European Communities), 2004a. Towards a Thematic Strategy on the Urban Environment, Communication from the Commission to the Council, the European Parliament, the European Economic and Social Committee and the Committee of the Regions (COM 2004(60) final), available on the Internet at: http://eur-lex.europa.eu/LexUriServ/LexUriServ.do? uri=COM:2004:0060:FIN:EN:PDF (accessed 19.07.10). Cherchi, E., 2009. Modelling individual preferences, state of the art, recent advances and future directions. Resource paper prepared for the Workshop on “Methodological Developments in Activity-Travel Behavior Analysis” at the 12th International Conference on Travel Behaviour Research held in Jaipur, India, 13–18 December 2009. Available on the Internet at http://iatbr2009.asu.edu/ ocs/custom/resource/W5 R1 Modelling%20individual%20preferences,% 20State%20of%20the%20art.pdf (accessed 19.07.10). Chrisman, M., 2002. Exploring Geographic Information Systems. John Wiley & Sons, New York, NY, United States. Commission of the European Communities (Directorate-General Regional Policy), 2008. Fostering the Urban Dimension. Analysis of the Operational Programmes Co-financed by the European Regional Development Fund (2007–2013), Working Document available on the Internet at http://ec.europa.eu/ regional policy/sources/docoffic/2007/working/urban dimension en.pdf (accessed 19.07.10). Environmental Systems Research Institute, 1999. Getting to Know ArcView®GIS. The Geographic Information System for Everyone. ESRI, Redlands, CA, United States. European Commission, Committee on Spatial Development, 1999. ESDP. European Spatial Development Perspective. Towards Balanced and Sustainable Development of the Territory of the European Union, Agreed at the Informal Council of Ministers Responsible for Spatial Planning in Potsdam, Germany, May 1999, Available on the Internet at http://ec.europa.eu/ regional policy/sources/docoffic/official/reports/pdf/sum en.pdf (accessed 19.07.10). European Commission, 2010. Europe 2020. A Strategy for Smart, Sustainable and Inclusive Growth, Communication from the Commission available on the Internet at http://ec.europa.eu/eu2020/pdf/ COMPLET%20EN%20BARROSO%20%20%20007%20-%20Europe%202020%20%20EN%20version.pdf (accessed 19.07.10). Greene, W.H., 1993. Econometric Analysis. Macmillan, New York, NY, United States. Hosmer, D.W., Lemeshow, S., 1989. Applied Logistic Regression. John Wiley & Sons, New York, NY, United States. Hospers, G.J., 2003. Localization in Europe’s periphery: tourism development in Sardinia. European Planning Studies 11 (6), 629–645. Hutchinson, S., 2004. Inside ArcView GIS 8.3. Thomson/Delmar Learning, Clifton Park, NY, United States.

485

Korte, G.B., 2001. The GIS Book, 5th ed. OnWord Press, Albany, NY, United States. McFadden, D., 1974a. Conditional logit analysis of quantitative choice behavior. In: Zarembka, P. (Ed.), Frontier of Econometrics. Academic Press, New York, NY, United States, pp. 105–142. McFadden, D., 1974b. The measurement of urban travel demand. Journal of Public Economics 3, 303–328. McFadden, D., 1976. The revealed preferences of a government bureaucracy: empirical evidence. Bell Journal of Economics and Management Science 7, 55–72. McFadden, D., 1978. Modelling the choice of residential location. In: Karlqvist, A., Lundqvist, L., Snickars, F., Weibull, J.W. (Eds.), Spatial Interaction Theory and Planning Models. North Holland, Amsterdam, The Netherlands, pp. 75–96. McFadden, D., 1980. Econometric models for probabilistic choice among products. The Journal of Business 53, 13–29. McFadden, D., 2000. Disaggregate behavioral travel demand’s RUM side. A 30-year retrospective. In: Hensher, D. (Ed.), Travel Behaviour Research: The Leading Edge. Pergamon Press, Oxford, United Kingdom, pp. 17–63. McFadden, D., Train, K., 1978. The goods/leisure tradeoff and disaggregate work trip mode choice models. Transportation Research 12, 349–353. McFadden, D., Train, K., Ben-Akiva, M., 1987. The demand for local telephone service: a fully discrete model of residential calling patterns and service choices. RAND Journal of Economics 18, 109–123. Nerlove, M., Press, S., 1973. Univariate and Multivariate Log-Linear and Logistic Models. RAND Corporation, Santa Monica, CA, United States, Report No. R1306EDA/NIH. Ortúzar, J. de D., Willumsen, L.G., 2001. Modelling Transport, 3rd ed. John Wiley & Sons, Chichester, United Kingdom. Pavlopoulos, D., Muffels, R., Vermunt, J.K., 2010. Wage mobility in Europe: a comparative analysis using restricted multinomial logit regression. Quality and Quantity 44, 115–129. Regione Autonoma della Sardegna (Autonomous Regional Administration of Sardinia), 2006a. POR Sardegna 2000–2006 (2000–2006 Sardinian Regional Operational Program), Available on the Internet at http://www.regione. sardegna.it/documenti/1 84 20080409115630.zip (accessed 19.07.10). Regione Autonoma della Sardegna (Assessorato della Programmazione, Bilancio, Credito e Assetto del Territorio, Centro Regionale di Programmazione, Laboratorio Territoriale della Provincia di Cagliari) (Autonomous Regional Administration of Sardinia (Office for Programming, Budget, Credit and Territorial Organization, Regional Programming Center, Territorial Workshop of the Province of Cagliari)), 2006b. Progettazione Integrata. Rapporto d’area – Provincia di Cagliari (Integrated Projects. Local Report – Province of Cagliari), Available on the Internet at http://www.regione.sardegna.it/documenti/1 85 20060531162844.zip (accessed 19.07.10). Regione Autonoma della Sardegna (Autonomous Regional Administration of Sardinia), 2007a. Programma operativo regionale Sardegna. Competitività regionale e occupazione (2007–2013 Sardinian Regional Operational Program. Regional Competition and Employment), Available on the Internet at http://www.regione.sardegna.it/documenti/1 84 20080219093313.pdf (accessed 19.07.10). Regione Autonoma della Sardegna (Autonomous Regional Administration of Sardinia), 2007b. Complemento di Programmazione del POR Sardegna 2000–2006. Approvato dal Comitato di Sorveglianza con procedura scritta e notificato alla Commissione Europea il 13 dicembre 2007 (Programming Complement of the 2000–2006 ROP Sardinia. Approved by the Monitoring Committee through a written procedure and notified to the European Commission on December 13, 2007), Document and annexes available on the Internet at http://www.regione.sardegna.it/documenti/1 84 20080118131426.pdf and http://www.regione.sardegna.it/documenti/1 84 20080118130438.zip (accessed 19.07.10). Train, K., 2009. Discrete Choice Methods with Simulation, 2nd ed. Cambridge University Press, Cambridge, United Kingdom. Tversky, A., 1972. Elimination by aspects: a theory of choice. Psychological Review 79, 281–299. Williams, H.C.W.L., 1977. On the formation of travel demand models and economic evaluation measures of user benefit. Environment and Planning A 9, 285–344. Zoppi, C, 2008. Questioni conflittuali nel processo di adeguamento del Piano urbanistico del Comune di Sinnai al Piano paesaggistico regionale: un caso di studio (Conflicts in the adjustment process to the Regional landscape plan of the Masterplan of the City of Sinnai: a case study). In: Zoppi, C. (Ed.), Governance, Pianificazione e Valutazione Strategica (Governance, Planning and Strategic Assessment). Gangemi, Rome, Italy, pp. 329–369.