Abstract
Local economic development is a much used tool for the regeneration of urban areas. Although the designation and effectiveness of local economic development policies has been studied extensively in existing literature, the question as to whether these policies are aimed at the areas that are most in need of these policies remains relatively understudied. This question has been answered using the case of industrial sites in the Netherlands. This particular type of urban area in the Netherlands has experienced problems with rapid urban area decline and has therefore been targeted by various area-based regeneration initiatives for many years. The economic performance is the main justification for the designation of industrial sites that are in greater need of regeneration. However, another pertinent and unanswered question is: are the industrial sites targeted for regeneration really the ones that underperform economically? (Multinomial) logistic regression analysis is used to answer this question. Differences in economic performance, measured by the growth in employment figures, the number of companies and property values, between industrial sites that are targeted for regeneration in two different rounds of regeneration initiatives and non-targeted sites are studied. The analysis shows that the differences in economic performance are negligible between these groups. It indicates that other criteria, such as political and strategic decision making influence policymakers’ decisions to target industrial sites for regeneration, thus making it at least doubtful that public money for the regeneration of industrial sites is being spent on what it is meant for.
Keywords
Introduction
The term decline ‘in the context of the urban development is used to describe undesirable changes, such as job losses accompanied by growing unemployment, social exclusion, physical decay and worsening living conditions’ (Lang, 2005: 2). The focus of this study is on the decline at the level of urban areas, more specifically on the decline of industrial sites. 1 These areas are often targeted by national, regional or local government initiatives to counter the negative effects of decline. This is most frequently through policies that stimulate local economic development and finance physical improvements. The nature of these initiatives differ. In this paper we analyse, in greater detail, the studies on enterprise zones (EZs) and tax increment financing (TIF) as these are much used programmes by which governments have tried to regenerate industrial areas that are referred to as ‘blighted’ or ‘economically distressed’. Although problems may vary and definitions of urban blight, decay, decline and other terms overlap considerably, these all express processes similar to the ones mentioned by Lang (2005) above. From here, urban areas that experience such problems are referred to as ‘declined’. The problems reflecting decline are various and include for example a ‘fall in property values and general development activity’ (Turok, 1992: 361) and social problems related to ‘increases in inner-city unemployment and poverty’ (Broadway and Jesty, 1998: 1423). Area-based regeneration initiatives aim to counter these negative effects, for example through ‘promot[ing] economic development in specific locations, typically economically-distressed urban areas’ (Neumark and Kolko, 2010: 1), ‘fostering and facilitating market driven urban renewal’ (Jolin et al., 1998: 81) and ‘target[ing] various economic development incentives towards specific blighted areas’ (Greenbaum and Engberg, 2004: 315). These aims illustrate that the justification of area-based regeneration initiatives are often found in increasing the economic performance of urban areas. This is commonly done through local economic development which has led to a substantial amount of literature on the effectiveness of area-based regeneration initiatives such as EZs and TIF (see e.g. Bondonio and Engberg, 2000; Greenbaum and Bondonio, 2004; Greenbaum et al., 2010; Neumark and Kolko, 2010). The question as to whether regeneration initiatives actually target the neediest and most distressed areas has not been researched much (Greenbaum and Bondonio, 2004; Greenbaum et al., 2010). Different authors nevertheless stress the relevance of such a question for economic development policies (Anderson and Wassmer, 2000; Bartik, 1991, 1994; Peters and Fisher, 2002; Shridhar, 2001) so the question that we address in this paper is whether regeneration initiatives target the areas they actually intend to target.
This paper empirically examines whether regeneration initiatives for industrial sites are targeted at the industrial sites that are actually in need of regeneration. With the economic justification of these initiatives in mind, the areas that are targeted should be the ones that underperform on variables that are related to economic performance. The empirical part of this paper focuses on regeneration initiatives for declined industrial sites in the Netherlands. It consists of logistic regression analyses that are used to find possible differences between industrial sites that were targeted for regeneration and sites that were not. In the Netherlands, industrial sites have appeared firmly on the agenda of policymakers due to problems with rapid decline (van der Krabben and Buitelaar, 2011). This has led to large government funded regeneration initiatives for industrial sites. These sites consist mainly of subsidies for physical improvements of infrastructure, the public and private space. The exact nature of the regeneration initiatives studied in this paper are explained in more detail below. Research by Ploegmakers and Beckers (2015) indicates that these have a negligible impact on various indicators of local economic development, such as growth in employment, number of firms and property values, leading to the question of whether these policies actually target the areas they intend to.
The remainder of this paper is set up as follows. In the next section we elaborate on existing research on area-based regeneration initiatives and focus on the question of whether local economic development policies are aimed at the areas they aspire to target. Furthermore, the process of decline of industrial sites, the rationale behind it, and the nature of current regeneration initiatives in the Netherlands are explained in more detail. The third section provides the description of the data that were used to perform logistic regression and multinomial logistic regression, the results of which are presented in the fourth section. The final section concludes.
Targeting industrial sites for regeneration
Introduction
Area-based regeneration initiatives are an important means to stimulate local economic development in declining urban areas. Two different mechanisms through which local economic development typically is stimulated can be distinguished in literature: business incentives and physical improvements (see e.g. Bartik, 2004; Ho, 1999; Rhodes et al., 2005; Tyler et al., 2013). Examples of physical improvements include the provision of new infrastructure, investments in public space, refurbishments of derelict private land and other property, and so on. Under business incentives we distinguish between a variety of measures without a clear physical component, such as tax breaks for firms locating in an area, business loans, enterprise support and skills training for employees and businesses. Regeneration programmes for industrial sites in the Netherlands are publicly funded and consist almost entirely of financing physical improvements. Well-known types of area based initiatives in other countries that have also been studied extensively in urban literature, such as TIF and EZs, resemble Dutch regeneration programmes in the sense that they are also specifically aimed at certain areas. While physical improvements are an important aspect of TIF, EZ programmes are usually aimed at attracting new businesses or stimulating investments by firms in the targeted area (Sridhar, 2001). However, most programmes do also offer subsidies for improvements to infrastructure (Greenbaum and Engberg, 2004). This makes both types of area-based regeneration initiative interesting for the present paper. A substantial part of the studies on TIF and EZs focuses on questions of effectiveness by empirically evaluating different programmes for metropolitan areas, states or cities. For EZs these include, for example, Bondonio and Engberg (2000), Busso et al. (2010), Greenbaum and Engberg (2004), Neumark and Kolko (2010) and Sridhar (2001). Examples of studies that have focused on the effectiveness of TIF include Byrne (2010), Dye and Merriman (2000), Lester (2014), Man (1998), Man and Rosentraub (1998) and Weber et al. (2003). As mentioned earlier, our aim is not to add yet another evaluation of a certain area-based regeneration programme. Instead, we focus on the question of whether an existing regeneration programme does indeed target the areas that are in need of regeneration. Existing evaluations, however, may benefit from the insights this research provides, as it is difficult to evaluate whether regeneration programmes do succeed, without first knowing whether these programmes do, in fact, target the areas they were intended to target. Below, we elaborate studies on TIF and EZs. Although most studies focus on effects and evaluations, some studies also pay attention to the justification of the programmes they study and elaborate, for example, which municipalities adopt TIF or what the characteristics of designated EZ districts are. TIF and EZs do bear some similarities but are not exactly the same as Dutch regeneration programmes; it is especially these issues we are interested in and therefore the exact nature of the programmes is less relevant. At the end of this section, the nature of regeneration programmes in the Netherlands is explained in more detail.
Targeting the intended areas
Many studies perform evaluations or focus on effectiveness of area-based regeneration initiatives, as illustrated above. There are also several authors that have focused explicitly on the importance of targeting the neediest areas. Bartik (1994) has assessed different state and local economic development programmes when developing advice on how these programmes could play a constructive role in the development of the urban economy. The focus is on programmes in which business incentives, or what he calls ‘customized business assistance’ programmes, play an important role. This type of local economic development programme covers many different policies, ranging from financial incentives such as property tax abatements to capital market policies such as government-financed loans for new businesses. Their main focus is to create employment. An important goal of these local economic development programmes is ‘to explicitly target economic development on distressed areas’ (Bartik, 1994: 17). Additional jobs in distressed areas can be more valuable than additional jobs in areas that already are performing well. In that way redistributing economic activity to distressed areas could increase the overall economic efficiency, although this may be difficult to realise deliberately through policy. In an earlier work Bartik (1991) argued that job creation is more efficient in distressed areas that already have typically high unemployment rates. There are several other authors that deliver the argument that local economic development policies should maintain focus in order to continue with their effectiveness. Peters and Fisher (2004) for example state that it is politically difficult to maintain focus for development programmes and that when, over time, targeting erodes, a wider range of localities are granted similar tools to compete with each other for new investments. Eventually, this will lead to a disadvantage for the older, more distressed areas such as greenfield sites and incentives are more readily given to smaller congested brownfield sites. Greenbaum et al. (2010) add to this, stating that from a policy perspective, it is important to keep focusing on how incentives should be targeted as political pressures can easily lead to the spread of programmes to less economically justified places over time, influencing programme effectiveness.
Wassmer and Anderson (2001) agree with Bartik (1994) that especially blighted areas should be targeted. The empirical results of their study of locally offered development incentives to attract businesses in the Detroit area leads to the conclusion that targeting the most blighted areas is preferable over two other possible options: a non-restricted use of local incentives by local policymakers or to ban all local development incentives. The first may lead to efficiency problems since ‘as time passes communities are more likely to emulate each other and offer manufacturing abatements just because others are using them’ (Wassmer and Anderson, 2001). If this is indeed the case, eliminating all local economic development policies might be an option to consider, but the authors prefer targeted use of local incentives for areas that are blighted, which they define in terms of employment figures and property values. This preference, although not specified by the authors in more detail, is shared by Dutch policymakers when it comes to local economic development, as will be elaborated on in the following section. Sridhar (2001) draws a similar conclusion in her benefit-cost analysis of a regional development programme, the Ohio Enterprise Zone Program. Her analysis shows that the creation of jobs will be more efficient in areas with high levels of unemployment. The use of more selective designation criteria could also be helpful as it reduces the competition between zones. It will give a clearer distinction as to which areas are eligible to become an EZ and which areas are not. She therefore argues that the selection of areas that are eligible for tax incentives aimed at regional economic development should be based on a thorough assessment of distress criteria. In their review of local economic development policies, Peters and Fisher (2004) thus conclude that, when effectively aimed at poorer areas, targeted programmes are the best that the economic development industry has to offer. Greenbaum et al. (2010) have empirically studied the geographic distribution of economic development tax incentives in Ohio and have found that ‘policies are missing the mark if they are indeed intended to target areas of distress or particular industries’ (Greenbaum et al., 2010: 154).
Many studies on TIF focus on effectiveness (see e.g. Anderson, 1990; Dye and Merriman, 2000; Lester, 2014; Weber et al., 2003). An important difference between these studies and the present paper is the explicit focus on the question of what types of areas are targeted by area-based initiatives and whether these are indeed those that are intended to be targeted. This is clearly illustrated by the fact that in many of the existing studies mentioned above, probability scores for characteristics of the targeted areas are estimated using probit or multinomial regression analysis, only to control for sample selection bias, as a first step in evaluating the effects of local economic development policies. In this paper we estimate similar probabilities, but explicitly focus on these scores as the most important outcome to conclude on what types of areas are targeted.
Studies on EZs have been more explicit in their focus on the question of what areas are targeted. For local economic development programmes that target distressed areas, these distressed areas – by definition – are likely to have higher levels of economic distress than non-designated areas (for evidence, see Bondonio and Engberg, 2000; Bondonio and Greenbaum, 2007; Greenbaum and Bondonio, 2004). Therefore, any study that compares the levels of economic outcomes for targeted areas versus some comparison group of areas is likely to be biased towards finding negative effects of the programme. This is due to the levels of economic outcomes which are obviously positively correlated over time, and therefore the targeted areas would have higher levels of distress than their comparison group in the future without the programme’s intervention. It is not as obvious that changes in economic outcomes will differ between targeted areas and comparison non-targeted areas. For targeted areas, a number of studies in the United States have attempted to evaluate EZs by comparing the performance of EZs to matched non-zone areas. Several studies, including the ones mentioned above, have explicitly made such matches using estimates of the ‘propensity score’. That is, estimates of the probability of a given area (in this case, a postal ‘zipcode’ or routing code) being designated as an EZ. These studies find little or no effect of EZ designation on the growth of local business activity. In addition, as mentioned before, the propensity score estimation suggests that EZ designation is not strongly correlated with previous area growth, which increases the odds that the estimates reveal the true effect of EZ designation (Bartik, 2004).
Two important assumptions made from the studies elaborated above also apply to Dutch regeneration initiatives. Firstly, stimulating employment leads to economic growth (economic growth is often measured in terms of employment growth). Secondly, in order for local economic development policies to be effective, it is important to target areas that are really in need of these policies. Based on the justification used for the policies and the findings from the empirical studies above, the areas that are intended to be targeted are the areas that experience the largest decline-related problems and those which are underperforming economically. Nevertheless, the above review of empirical studies shows that political and strategic reasons should also be taken into account as a possible explanation as to why an industrial site was designated as one in need of regeneration while other sites are more eligible to be targeted according to the original aim of regeneration programmes. This leads to the hypothesis that the industrial sites that are actually targeted for regeneration initiatives are not necessarily industrial sites that underperform in terms of the variables that represent the distress criteria mentioned above. In fact, van den Anker et al. (2009) note that political and strategic reasons are instrumental in the decision to target industrial sites for regeneration: ‘If grants are available, not much effort is needed to designate a site as in need of regeneration’ (van den Anker et al., 2009: 52 (translation by authors)). In the next section decline and regeneration of industrial sites in the Netherlands are elaborated on to give a more detailed explanation.
Decline and regeneration of industrial sites in the Netherlands
In their work on industrial sites in the Netherlands, Louw et al. (2009) distinguish five categories of problems that are related to the decline of industrial sites. These are problems regarding private property, appearance, accessibility and infrastructure, (inefficient) use of space, and environmental issues (ETIN-Adviseurs, 2003; Louw et al., 2009). The decline of individual private property plays an important role in the overall decline of industrial sites in the Netherlands (and elsewhere). As time passes, all property is subject to decline. This process of decline can go faster under the influence of decreasing accessibility, the diminishing quality of public space and increasing unsafety, since these may lead to disinvestments by owners of private property. In the case of industrial sites in the Netherlands, the abundant supply of new industrial sites at relatively low prices has led to further disinvestments in the existing stock of industrial property (Planbureau voor de Leefomgeving, 2009b; van der Krabben and Buitelaar, 2011). In a wider context, Healey (1991: 100) has also made the link between urban decline and the condition of property:
the conventional economic assumption is that if the overall level of economic activity in a place falls, the amount of wealth to invest in property and the quantity of demand falls. As a result, land and property values fall. This discourages further investment both in new development and in maintenance of the existing stock.
The physical appearance of industrial sites is strongly linked to the decline of private property, which makes up an important part of any industrial site. Physical appearance is also closely connected to the public infrastructure and public space of an industrial site. Government investments are usually aimed at the quality of the public space and infrastructure. This is almost exclusively provided publicly and can also have an important impact on the appearance. The effect on the investment behaviour of private property owners of public investment is potentially large. The justification for public investments at least seems to rely, for a large part, on this relationship (Ploegmakers and Beckers, 2015). Public investments are not only a goal, but are seen as a means through which private property owners can be persuaded to invest in not only their property, but also in new economic activities.
The third category of problems concerns accessibility. Each firm has different demands in terms of accessibility: some companies should be able to easily transport heavy goods over water or via rail, whereas for others easy car accessibility and abundant parking for customers and employees is essential. Many of the industrial sites in the Netherlands which experience decline, face problems with poor infrastructure and accessibility. Demands regarding infrastructure may change over time, for instance when it comes to the demands regarding the framework for information technology purposes. Industrial sites should therefore not only offer physical, but also the right digital infrastructure.
Fourth, inefficient use of space could be an indication of the decline of industrial sites in the Netherlands. In many urbanised areas, (open) space is a valuable asset and high vacancy rates and inefficient lay-outs of existing industrial sites are therefore regarded as undesirable. Both inefficient use of space and vacancies in a certain part of the site can have a negative effect on the rest of the industrial site, causing a negative spiral and leading to decline.
The final category concerns environmental issues, taking into account not only contamination of soil, but also nuisance created by smoke, fumes or noise. Furthermore, the presence of companies that pose a risk to the environment, for instance due to their production of chemicals or the attraction of large volumes of transport, might cause an industrial site to decline. This is most notably because it makes the location of a site obsolete, when residential areas located nearby have been developed.
Normally industrial sites are targeted for regeneration by municipalities when these areas are faced with problems elaborated above. Typical aspects of regeneration programmes include physical improvements of infrastructure, the public and private space. The latter may include relocating undesired activities, acquiring and demolishing obsolete property providing new building land for firms looking to locate in redeveloped parts of an industrial site. Firms are also stimulated to form co-operations to manage the public space. The majority of these measures are funded through grant programmes from the national and provincial government. The funding schemes available range from broad urban regeneration programmes to grant funds especially for specific physical improvements such as infrastructure. There are special grants for soil remediation in case of contamination but this is only the case for a relatively small number of industrial sites (Ploegmakers and Beckers, 2015).
Ploegmakers and Beckers (2015) have studied the effects of regeneration initiatives for industrial sites in the Netherlands, but in their study the issue as to whether these initiatives are aimed at the industrial sites that should be targeted is not addressed. They study municipal master plans to find the most frequently cited goals of regeneration initiatives. Two frequently cited goals are ‘attraction and retention of firms’ and ‘increasing the number of jobs’, indicating that the assumption that employment leads to economic growth, presented earlier, underlies Dutch regeneration policies. The goals of regeneration programmes are in line with the categories distinguished by Louw et al. (2009) elaborated above. Regeneration initiatives aim to generate positive effects on one or more of these categories. In the next section the dataset is introduced that is used to answer the question of whether regeneration initiatives indeed focus on the industrial sites most in need of regeneration.
Data description
In this section the dataset for the empirical study is presented and the explanatory variables are introduced. Two logistic regression analyses are performed. First, standard logistic regression analysis is used to find if the probabilities for characteristics of industrial sites differ between industrial sites that were targeted for regeneration and non-targeted sites. The second, multinomial logistic regression analysis, allows for a more detailed analysis of the differences in characteristics within the group of targeted industrial sites. Although we make use of a longitudinal dataset, the nature of the dependent variable only permits an analysis of cross-sectional variation in the decision to target a particular area for regeneration (we do make use of the panel characteristics of the dataset to calculate changes in the main variables of interest over a longer period of time (1997–2008) preceding the decision to target the area). Clearly, it would be preferable to undertake a longitudinal analysis. However, we are able to add a time dimension to the data. In 2009 the national government and lower-tier governments agreed on the future direction of regeneration policies. This agreement distinguished two rounds of regeneration. The first round, which would be implemented between 2009 and 2013, was targeted at 6500 hectares of industrial sites experiencing decline. The second round would target 9300 hectares and had to be finished in 2020.
The information on which sites had to be targeted during the first round comes from lists provided by the provinces. Provinces compiled these lists together with municipalities. They are based on municipal knowledge and, in many regions, private consultancy firms were hired to make a general assessment of the condition of all industrial sites. Unfortunately, such lists are not available for the second round of the regeneration programme foreseen to be implemented in the period 2014–2020. Therefore, we rely on the Dutch national database on industrial sites (IBIS), which tracks the amount of land in need of regeneration on an annual basis. Sites are assigned to the group of second round sites if they were suffering from some form of decline 2 according to the 2008 annual IBIS assessment, but not targeted for regeneration during the first round.
The characteristics of industrial sites are the explanatory variables in our dataset. Different existing databases were used to construct this dataset. IBIS contains data on size, land prices and location among others, and was combined with data on property values obtained from Netherlands Statistics (CBS) 3 and data gathered in cooperation with the Netherlands Environmental Assessment Agency (PBL). 4 This provided the opportunity to construct a rich database on all industrial sites in the Netherlands.
Research shows that municipalities to a large extent apply the same criteria to decide whether an industrial site faces decline and if it is indeed in need of regeneration (for these criteria see Planbureau voor de Leefomgeving, 2009a; Ploegmakers and Beckers, 2015). The criteria and guidelines set at the national level provide certainty that the decision by policymakers to target an industrial site is made on comparable grounds and should not differ substantially from one (municipal or regional) policymaker to another. Our analysis is thus based on the assumption that the most-declined industrial sites are targeted. Municipalities and regional government however can have other reasons to target industrial sites within their jurisdiction, or set their own priorities based on how many industrial sites have declined. Municipalities with no industrial sites in decline might still target such areas to try to receive subsidies. To control regional differences in interpretation of the national guidelines, provincial dummies were included in the analyses. 5
Two types of explanatory variables are distinguished: control variables and performance variables. The first category is divided into site-specific characteristics and regional characteristics. These variables might influence the probability that a certain industrial site is targeted for regeneration. They are mostly static and are not used as justification for regeneration policies. Examples of such characteristics are age, travel time to/from the nearest motorway exit and urbanisation rate. Older industrial sites are probably more likely to be targeted for regeneration. This, however, is a characteristic which is impossible for policymakers to change. The influence of these variables is controlled for by including these in the analysis. The second category corresponds to the economic distress criteria. These are called performance characteristics as they express the economic performance of the industrial site before interventions are implemented. Different indicators of economic activity can be found in studies that were reviewed in the previous section. Employment figures and the number of companies in the targeted area are central in almost every study on local economic development. Wassmer and Anderson (2001) further take into account property taxes paid (as an indicator for property values) as a measure of economic activity. These criteria correspond with negative developments that the regeneration initiatives typically aim to counter: loss of jobs and a diminishing tax base. Following Ploegmakers and Beckers (2015), growth of jobs, growth of the number of companies and the growth of property values are used as these are important indicators of economic performance which, in turn, is the main justification for regeneration initiatives of industrial sites in the Netherlands.
The explanatory variables and the expected relationship with the dependent variable included in the analyses are the following. Travel time to/from the nearest motorway exit represents accessibility. Accessibility is better for industrial sites for which travel time to/from a motorway is lower and these sites are therefore less likely to be targeted. The variables of residential land use and open space land use are included in the model to control for functions in the direct surrounding of the industrial site. Residential neighbourhoods in the vicinity of an industrial site may lead to nuisance and environmental hazards, positively affecting the chance of being targeted. Following this line of reasoning, a negative relation is expected between open space and targeting. Age is included due to the fact that older industrial sites are hypothetically more likely to be targeted. The variable type of industrial site is included to control for the influence of the presence of certain types of firms. 6 Industrial sites with a large share of manufacturing could be more likely to experience decline and be targeted for regeneration. One reason for this is because firms in this sector may have recently seen a down turn in employment. Number of jobs and number of companies represent the size of industrial sites to control for large and smaller industrial sites. Jobs per hectare is included in the model as the representation of efficiency of land use. As described above, inefficient use of land is one of the categories of problem that is related to decline. It is expected that municipalities consider industrial sites with small numbers of jobs per hectare as less efficient in terms of land use. Increasing the efficient use of land is often mentioned as a goal of regeneration policies and the number of jobs per hectare therefore is expected to influence the municipal decision to target an industrial site for regeneration. Environmental classes are included to control for the presence of companies that have a potentially large effect on environmental issues, including noise, fumes or external hazards. This often concerns large industrial installations or (chemical) plants and these companies are not easily moved to another location. Their present location on an industrial site should therefore be maintained in the best possible way and regeneration is used to guarantee this. A second possible option is that these companies cause more environmental issues and this increases the need for regeneration initiatives to counter these negative effects. Seaports are included to control for this specific type of industrial site. These types of industrial sites typically have a low number of jobs per hectare. Average property values are added to control for low quality public space and private property. A low property value per hectare is an indication of low overall quality and therefore might influence the assessment of targeting. Three regional characteristics are added. Urbanisation rate is included to control for possible differences between urbanised milieus industrial sites can be located in. Three categories are distinguished: urban agglomeration, city region and outside city region. Within urban agglomeration the necessity to regenerate existing industrial sites is expected to be higher than in more rural municipalities outside city regions. This can be attributed to the pressure on maintaining the quality of the built up space being higher in more urbanised areas. Scarcity is calculated as the ratio of readily available and land currently in use for industrial sites. This variable is included to investigate the often-suggested relation between the (cheap) supply of new land for industrial sites and the occurrence of decline and the subsequent need for regeneration. Although not studied empirically, it is thought that when new land for industrial sites is cheap and easily available, firms are less willing to invest in maintenance at their present location. A high ratio means high levels of scarcity and a negative relation between this variable and targeting is therefore expected. Finally, we have added a regional (provincial) dummy to control for provincial policymakers’ decisions to target industrial sites for regeneration. 7 Table 1 lists the explanatory variables, their definitions and descriptive statistics along with the results of analyses of the variances of means. The results show there are noticeable and significant differences between the mean values of targeted and non-targeted industrial sites. Two of the performance characteristics have significantly different mean values. The mean differences between round 1 and round 2 targeted sites show fewer differences. The regional dummies show large significant differences, indicating that the number of industrial sites targeted differs per province.
Descriptive statistics of explanatory variables in dataset and t-test results for differences in variance of mean values between targeted and non-targeted sites, and round 1 and round 2 targeted sites.
Note: Tests of the equality of the means: **significant at the 1% level; *significant at the 5% level.
The group of targeted (round 1 and round 2) sites consists of 905 sites (of which 446 are targeted in round 1) while there are 817 non-targeted sites in the total dataset of 1722 industrial sites. From the mean values for site characteristics it becomes clear that, on average, the problem sites are older. Almost 40% of the targeted sites are from the pre-1960s period, whereas the largest group of non-targeted sites is from the 1980s. Interestingly, 5% of targeted and declined sites were developed in the 1990s and are thus relatively young to already be experiencing problems related to decline. Industrial sites dominated by manufacturing make up the largest group for both targeted and non-targeted sites. Mixed use is the second largest for both categories, although the group of mixed use sites represents 39% of the targeted and only 25% of non-targeted sites. Targeted sites are bigger, as measured by the total number of jobs and number of firms. The number of jobs per hectare however is higher for non-targeted sites. Average property values are 11% higher on non-targeted sites. The regional characteristics do not show large differences. The industrial sites in the dataset are quite evenly spread among the regions and areas of urbanisation distinguished. The mean values for the provincial dummies show the distribution of industrial sites over the Netherlands. Large, traditionally industrialised provinces such as Noord Brabant and Gelderland house many industrial sites. The large cities Rotterdam and The Hague (Zuid Holland) and Amsterdam (Noord Holland) also have high shares of industrial sites.
The values of the performance characteristics in Table 1 indicate that targeted industrial sites show slower growth than non-targeted sites. Interestingly, the change in property value is higher for industrial sites that are targeted. Two further analyses of these differences in which the influence of the other variables is controlled for, are presented in the next section. In the first analysis it is expected that the results will show that a negative growth of the performance indicators will increase the chance of being targeted for regeneration. In the second, multinomial logistic regression analysis it is expected that all three of these variables will show a similar pattern: industrial sites that are targeted in the second round show negative results compared to the group of non-targeted sites. Industrial sites that are targeted for regeneration in the first round are expected to show more negative results compared to the industrial sites that are targeted in the second round.
Results: Differences between targeted and non-targeted sites
The results of the logistic regression analysis are presented in Table 2. 8 Non-targeted sites are the reference category. The results presented in Table 2 are interpreted to conclude on the probability that an industrial site belongs to the group of targeted sites. Travel time does not show a significant result. Residential land use does not influence the chance of being targeted, other than open space, which negatively influences the probability of being targeted. The categories of age again show the expected results: older industrial sites are more likely to be targeted than younger ones, with industrial sites that were developed before the 1960s showing the largest coefficient as compared to all other age cohorts and the reference category of the 1990s. An expected result is the non-significant coefficients for types of industrial sites. Compared to the reference category of mixed-use industrial sites, no type of industrial site has a significantly larger probability of being targeted to another. One possible interpretation is that all industrial sites are evenly prone to targeting. The same can be concluded for size, as the coefficients for number of jobs and number of firms does not show significant results. The coefficients for environmental impact classes indicate that industrial sites that house potentially polluting or hazardous companies are more likely to be targeted. The variable for property values also shows the expected negative coefficient. Under regional characteristics, urbanisation rate does not show significant results. Scarcity unexpectedly shows a positive sign, indicating that the often-suggested relation between readily available land and decline does not exist for our dataset. An alternative explanation is that high levels of scarcity might lead to a larger policy effort to regenerate industrial sites in the existing built up space and thus targeting occurs more often.
Results of the logistic regression. Probabilities for site characteristics, regional characteristics and performance characteristics on targeted sites (reference category: non-targeted sites).
Notes: Tests of the equality of the means: **significant at the 1% level; *significant at the 5% level.
Z-values in parentheses; results for provincial dummies are not reported for brevity.
The results for the performance characteristics do not show any significant relations between these variables and the probability of being targeted for regeneration. The coefficients for all variables are insignificant, indicating that, having controlled for the other variables in the model, there appears to be no significant relation between economic performance and targeting.
The second analysis is performed to find differences within the group of targeted industrial sites and compare these two groups with non-targeted sites. Hypothetically, and according to the theory laid out by Bartik (1994) among others, there are differences to be found for the performance indicators within the group of targeted sites as industrial sites targeted in round 1 are assessed as in need of regeneration before the sites targeted in round 2.
Table 3 denotes the results of the multinomial logistic regression analysis. Dividing the dataset into three categories does not change the overall pattern for most control variables. Other than expected, more travel time does not appear to be a reason to target industrial sites for regeneration. Round 1 industrial sites even show a positive coefficient for this variable. The coefficient for housing remains insignificant, while open space shows the expected negative sign for targeting in both rounds. Again, most age variables show the expected pattern, with only industrial sites developed in the 1980s showing non-significant results for the probability of being targeted in round 2 compared to industrial sites developed in the 1990s. A possible explanation for this is that sites in both these age bands belong to a similar ‘generation’ of industrial sites, developed in relatively the same style and with comparable characteristics (Louw et al., 2009; Olden, 2010). The presence of companies in the highest environmental impact class does not increase the probability to be targeted in round 2, but shows a significantly higher probability of being targeted in round 1. Overall, these results indicate that industrial sites that can house firms with a potentially large impact on the environment have a higher probability of being regenerated, as expected. Two possible explanations for the results found are the nuisance these types of companies cause to their surroundings, or the economic importance of these companies at their present location. This leads to a willingness for policymakers to invest in their present location to be able to accommodate economically important firms that are hard to relocate. Size has limited influence on targeting, with only round 2 sites showing a small positive relation with the number of jobs. The coefficient for property value per hectare shows a significant negative relation with the probability of being targeted in round 1. A low spatial quality, as represented by a low property value, thus leads to a prioritisation of being targeted. For scarcity, a negative coefficient with the probability for being targeted was expected, as this means a higher level of scarcity leads to less decline and subsequently, less regeneration is needed. For round 2 targeted sites, the opposite, unexpected effect is found. It appears that in the long run, scarcity leads to more attention for the regeneration of existing industrial sites. Urbanisation rates do not show significant results, indicating that regeneration occurs in both rural and urban municipalities alike.
Results of the multinomial regression analysis. Probabilities for site characteristics, regional characteristics and performance characteristics on round 1 and round 2 targeted industrial sites (reference category: non-targeted sites).
Notes: Tests of the equality of the means: **significant at the 1% level; *significant at the 5% level.
Z-values in parentheses; results for provincial dummies are not reported for brevity.
When it comes to the performance characteristics the results do not differ from the first analysis. None of the performance characteristics show significant results, indicating there are no differences between industrial sites targeted for regeneration in round 1, round 2 and non-targeted industrial sites for the change in the number of companies, jobs and industrial property values.
Overall, there is minimal statistically significant evidence that industrial sites that are targeted for regeneration in either round 1 or 2 show slower growth on performance characteristics than industrial sites that are not targeted. Furthermore, the differences between round 1 and round 2 targeting sites are also limited. Industrial sites targeted in round 1 do not necessarily appear to be in more need of regeneration, as measured by the performance characteristics, than sites targeted in round 2 and even non-targeted sites.
Conclusions
The question of whether the urban areas that are targeted for regeneration initiatives are actually the areas that are those most in need of regeneration has not been addressed much empirically in existing studies. Regeneration initiatives are often aimed at local economic development of blighted areas, but are the targeted areas indeed areas that suffer from decline? This paper set out to study the case of targeting rapidly declining industrial sites in the Netherlands to fill this void in existing literature. The results of the analyses performed in this paper provide an insight into the characteristics of industrial sites that are targeted for regeneration initiatives as opposed to sites that are not targeted. Furthermore, differences between sites that are targeted in two different rounds of regeneration programmes are studied in more detail. According to the justification that is used for regeneration initiatives, targeted industrial sites should be the ones that underperform in terms of economic performance. Economic performance is represented by three characteristics: growth of jobs, growth of number of companies and growth of property values. Theoretically, low values of the performance characteristics add to the probability of industrial sites being targeted for regeneration. High values add to the probability of falling into the category of non-targeted industrial sites. Industrial sites targeted in round 1 are prioritised, as these sites were subject to regeneration in the period 2011–2014. Sites targeted in round 2 are assessed as being in need of regeneration only after this period and are therefore expected to economically outperform round 1 sites.
The descriptive statistics show there are noticeable and significant differences between the mean values of targeted and non-targeted industrial sites, for example with respect to the age of industrial sites and environmental impact classes. The t-test results indicate lower average growth of number of jobs and growth of number of companies on targeted sites. These differences do not exist between industrial sites targeted in round 1 and round 2. The regional dummies show large significant differences, indicating the number of industrial sites targeted differs per province. However, the results of the logistic regression show that there are no differences between targeted and non-targeted sites. Dividing these into three categories does not change the pattern. Sites targeted for regeneration in round 1, round 2 and non-targeted sites do not show differences in economic performance. These results indicate that the economic performance of industrial sites bears little relevance on policymakers’ decisions to target industrial sites for regeneration. Therefore industrial sites that are targeted for regeneration are not necessarily the industrial sites that are the ones in need of these programmes. While local economic development policy is aimed at targeting sites that are struggling with the negative effects of decline, in practice targeted sites do not underperform on three of the characteristics that represent economic performance. Of course the results should be interpreted with care. The regional dummies included in the analysis control for possible differences in interpretation of decline by policymakers that can influence the targeting of industrial sites and may thus cause noise in the interpretation of our results. Despite the fact that guidelines for targeting are generally shared among policymakers, there can still be differences between the policymakers’ choices and this can possibly have an influence on the results of the empirical analysis. A possible alternative explanation for some of the results found therefore seems straightforward: strategic and political decision making. 9 Adding political and strategic decision making into a large scale analysis like the one presented in this paper is challenging and can be the subject of future research.
A second alternative explanation is elaborated here. The results might indicate that industrial sites are targeted at exactly the right moment in time: just before the industrial site is losing jobs and companies in numbers and the process of decline has led to disinvestments causing property to deteriorate and fall in value. In that case, policymakers may well assess and target the industrial sites they intend to target and measuring the performance of industrial sites actually fails in recognising these sites. From a policy perspective it seems appropriate to make investments before the problems that are related to decline have had a large influence on the industrial site. The investments to regenerate an already declined industrial site can be much larger and therefore less efficient. In fact Ploegmakers and Beckers (2015) mention that municipal master plans for regenerating industrial sites often cite that one of the goals is to prevent the sites that will, most likely, experience a downward spiral. However, one might wonder whether policymakers will be able to accurately identify this particular moment in time.
Another direction for future research is to further specify some of the variables used in this dataset. Type of industrial site does not show any significant results, although it is reasonable to expect differences in the probability of being targeted for different types of industrial sites. A more narrow specification for types of industrial sites might improve interpretation of results.
To conclude, our study clearly indicates that, for the Netherlands, it is at least doubtful that public money for the regeneration of industrial sites is always used for what it is meant for, namely targeting those industrial sites that are clearly underperforming in an economic sense compared to other sites.
Footnotes
Acknowledgements
The authors would like to thank three anonymous referees for their useful comments on earlier versions of this paper, and Jan Schuur, Hans van Amsterdam, and Marnix Breedijk at PBL Netherlands Environmental Assessment Agency for their support in analysing and making available the data.
Funding
This research was funded by a grant from the Netherlands Institute for City Innovation Studies (NICIS, currently Platform31).