Keywords

Introduction

Analyses of the URBAN I Initiative implementation in Spain between 1994 and 1999 have shown its adds value to urban policies promoting collaborative processes among different types of stakeholders in the design and implementation of urban policies (De Gregorio, 2015). However, there is no comparative analysis of its impact according to their objectives. Did the urban environment improve? Did economic activity in the neighbourhoods increase? Did the living conditions of residents improve?

As noted previously, analysis of the impact of these types of policies is scarce, and, in general, the outcomes are not conclusive, either because they do not exist, because they are usually moderate, or because they are different depending on the type of objective considered (Lawless, 2012; Navarro, Moya et al., 2016; Rhodes et al., 2005). This could be due to the difficulty of evaluating such policies because of their complexity, the temporal proximity between their implementation and impact analysis, or the lack of adequate data sources (see Chapter 1). However, as we pointed out already, it could also depend on the evaluation approach or strategy used. In this chapter, we propose a strategy that, using available data, considers some of these elements to analyse the impact of the URBAN I Initiative in Spain. We will point out the importance of the programme theory to analyse its impact, specifically, residents’ exposure to implemented strategies and the importance of addressing residential mobility as a factor explaining exposure to the programme. For this purpose, controlled comparisons will be applied to trajectories of change in urban areas between 1991 and 2001, considering the programme selection bias (eligibility criteria) and contextual exposure to the programme (as residential mobility).

The Impact of Integrated Urban Development and Urban Change: Intervention Mechanisms, Policy Exposure, and Residential Mobility

Integrated urban development projects use two major types of causal mechanisms to try to achieve their goals. On the one hand, mechanisms aim to produce contextual effects by improving the neighbourhood as a structure of opportunities for its residents and other agents that develop their activities in it (companies, associations, etc.). Here, it is assumed that there will be contextual effects on residents or other stakeholders resulting from their exposure to changes in the neighbourhood. On the other hand, mechanisms aim to improve the resources or lifestyles of specific groups through actions aimed at increasing their capacities regarding specific situations or behaviours (see Chapter 5).

Therefore, the project’s impact will differ depending on residents’ exposure to such mechanisms. Whereas the first type of mechanism could affect all residents because they are exposed to changes in the socio-spatial context (the neighbourhood), the second type would do so only, or fundamentally, for those exposed to—targeted by—specific actions. In part, the absence or low impact found in most area-based intervention evaluations might be explained because all residents are considered without regard to the degree or nature of their exposure to different policy actions included in projects. That is, they assume that projects are based exclusively on contextual mechanisms. This is partially due to the use of aggregate data in the absence of appropriate information at the individual level. In fact, when considering the degree and type of exposure to the project, we get a clearer picture of impacts that do not appear when this fact is not taken into account (Navarro, Rodríguez-García et al., 2016). This could be because the joint analysis of all residents produces composition effects between those most exposed to the programme, who might have been affected, and those not exposed to it, who are unlikely to have been affected.

Nevertheless, even assuming the contextual effect of the projects, be it due to a lack of specification in the project theory guiding the evaluation (assuming implicitly that its effects are only contextual), or because of the absence of appropriate data at the individual level, the evaluative analysis of such initiatives also faces the challenge of residential mobility. As in other neighbourhoods or urban areas, during project implementation, there is residential mobility, both outbound, because some of its residents move to other urban areas (outcomers), and inbound, because new residents move into the area (incomers). The volume and nature of this mobility can affect impact analysis. The few studies about residential mobility and area-based initiatives seem to show the existence of ‘upward residential mobility’ processes. Those whose situation improves as a result of the intervention tend to move to another neighbourhood and are replaced by other residents in worse socioeconomic situations (Cole et al., 2007; South et al., 2005). However, in other cases, interventions and their results may also attract new residents with a higher socioeconomic status. In these cases, neighbourhood improvements could be explained by population substitution more than an improvement in the lives of its traditional residents, as the literature on state-led gentrification points out (Hochstenbach, 2017). These two phenomena may underestimate (upward mobility) or overestimate (gentrification) project impact depending on the volume and characteristics of the ‘outcomers’ and ‘incomers’.

This means the impacts can be more clearly attributed to projects among those residents (or other stakeholders) who have remained in the neighbourhood throughout project implementation. Compared with incomers or outcomers, this group of stable residents (stayers) has been exposed during the whole implementation period; thus, in this case, changes could be more assuredly attributed to the project. Although their degree of exposure to specific targeted actions cannot be known, their contextual exposure to the project has been more intense than that of the other groups, albeit only for the duration thereof. It could, therefore, expect a more significant impact among them (Fig. 8.1).

Fig. 8.1
A block diagram represents the impact of pre-intervention, post-intervention, and contextual exposure, along with the impact probability.

The impact of integrated urban development strategies and residential mobility: the role of contextual policy exposure

Full size image

Research Design to Analyse Integrated Strategy Impacts: Controlling by Territorial Eligibility Criteria and Contextual Exposure to the Programme

As indicated in the previous chapter, to analyse the impact of integrated urban development programmes, it is necessary to conduct a controlled comparison between urban areas where these are applied and other areas that were similar before the implementation of projects. To this end, we applied the propensity score matching technique to select experimental and control urban areas according to the eligibility criteria established by the URBAN I programme. A total of 22 experimental areas and 98 control areas were selected (see Chapter 7). In addition to eligibility criteria (basically, urban vulnerability), we will also try to control the effect of policy contextual exposure due to residential mobility.

We analysed the patterns of change in these two types of areas for some indicators related to the programme objectives: business density (the number of establishments that develop an economic activity per 1000 residents), the rate of employment (the percentage of employed people over the total population of working age), the percentage of the university-educated population over the total adult population, and residents average socioeconomic level based on their occupations, from unskilled workers to executives and managers (scale values 0 and 1, respectively).Footnote 1

We have no individual longitudinal data for urban areas or information about residential mobility. Therefore, we can not differentiate the degree of contextual exposure to the programme at an individual level (either on the duration or because of the exposure to specific targeted actions). Therefore, we will approach this by performing two models with aggregated data with experimental and control urban areas as observation units and applying a repeat measures design. One model includes all residents in 1991 and 2001, the other only stable residents in 2001 in the post-intervention measurement. Therefore, in both models, the post-intervention measurement does not include those who have left the neighbourhood (outcomers). The difference is that the first model comprises both incomers and stayers, and the second contains only the latter. Therefore, the second model controls programme exposure derived from residential mobility, thereby providing a closer approximation of the impact of projects on residents. In addition, if there were any positive impacts on stable residents (second model), this would point to a possible revitalisation of the neighbourhood more than a led-state gentrification process, in other words, an improvement among traditional residents regardless of possible incomers with a higher socioeconomic position.

We have analysed the programme’s impact using the dRM indicator proposed by Morris and DeShon (2002). This measures whether the improvement trend in the experimental areas has been higher than in the control areas as standardised differences, so the impact in different indicators can be compared regardless of their original measurement scale.Footnote 2 For their interpretation, the basic rules outlined in the second chapter can be used: the effect size value, the confidence interval, and the percentage of experimental areas that show a trend of improvement above the average of the control areas.

We expect the impacts will be more evident, dRM values will be greater in models model controlling for contextual exposure—those including only stayers in the post-intervention measurement (Fig. 8.2). However, our proposal might also have certain limitations that should be pointed out. Firstly, it assumes that residential mobility patterns are similar in experimental and control areas in terms of volume and those involved. No differences exist in the rate of those who remain in the urban area (percentage of stayers in 2001), either for the population as a whole (which is 60%) or for five-year age groups, since life cycle is a significant factor in this phenomenon.Footnote 3 Secondly, we assume that stable residents are similar to the residential population in 1991; in other words, at the start of the intervention there were no significant differences between stable residents and those that moved to other areas before the end of the intervention. And thirdly, if the thesis of ‘upward residential mobility’ is true, we do not include the possible impact of the programme on outcomers, and therefore we might be underestimating it.

Fig. 8.2
Two graphs plot the impacts of d R M for all residents during the periods between 1991 and 2001, along with the stayers in 2001.

The logic of the quasi-experimental design for evaluating performance trends: repeat measures and control for policy contextual exposure

Full size image

Results

The change in the business density of urban areas analysed between 1991 and 2001 was 18 points, slightly higher in experimental areas than in control areas (22 and 17 points difference, respectively, see Table 8.1). This change shows the cycle of economic growth that took place during part of the analysis period, with a considerable increase in companies in Spain.Footnote 4 However, our analyses show that the trend is more prominent in the experimental areas, which could provide evidence of the impact of the URBAN Initiative.Footnote 5

Table 8.1 Changes in experimental and control urban areas between 1991 and 2001 (Difference 2001–1991. Mean [Standard deviation])
Full size table

Indicators for all residents show improvement patterns, although no programme impacts. The employment rate increases by 4 points, showing a pattern similar to the country as a whole during the period analysed (Jiménez et al., 2002), and the change is quite similar between experimental and control areas (4.6 and 4.2 points). The change in socioeconomic status is also low and similar in both types of areas (0.036 and 0.032 points, respectively). Finally, the percentage of population with a degree increases between 7 and 8 points in the analysed period, slightly higher in the experimental areas (7.7 and 8.1 points, respectively). Therefore, there do not appear to be any evident impacts from the URBAN initiative.

Analyses including only stable residents show the intensity of change between 1991 and 2001 is somewhat lower (Table 8.1). However, the differences between experimental and equivalent areas are greater than when considering all residents. Regarding the employment rate, this difference equals 0.42 when all residents are considered and 0.73 when only stable residents are examined. Concerning socioeconomic status, these differences are equal to 0.004 and 0.009, respectively, and equal to 0.467 and 0.803 for the university-educated population. These differences would show improvement patterns are clearer when models only include residents who remained in the neighbourhood between 1991 and 2001.

Figure 8.3 shows the impact of the URBAN I programme in terms of standardised differences between experimental and equivalent areas (dRM). Improvements in business density are greater in experimental areas than in control areas; the effect is moderate, although statistically significant (dRM = 0.336). The results would indicate that 59% of the experimental areas show a higher pattern of improvement than the average of the control areas. Therefore, although the effect is moderate, the URBAN I programme achieved its objective of increasing economic activity in the neighbourhoods. This does not mean that residents or existing businesses have set up this more intense economic activity. Policy actions might have promoted this, but also because they have made the neighbourhood a more attractive space for economic activity initiated by other external agents, as noted by analysis of similar initiatives (Archibald et al., 2019).

Fig. 8.3
A graph represents the data for business density along with employment rate, socioeconomic status, and Universitary of all and stayers.

The impact of the URBAN programme according to contextual programme exposure: all residents and stable residents (Effect size (dRM) and confidence intervals [CI90%])

Full size image

The impacts on the living conditions of all residents are very small or non-existent, at least for the three indicators considered here: employment rate (dRM = 0.147), average socioeconomic status (dRM = 0.126), and the university-educated population (dRM = 0.129). However, the evidence is clearer when the model includes only those residents exposed to the programme. Although the confidence interval would indicate that the differences are not statistically significant, the results concerning the employment rate show that 60% of the experimental areas improve above the control areas (dRM = 0.241). The impact is also moderate for the socioeconomic status of residents, as about 65% of the experimental areas have experienced a better pattern of change than the average improvement of the control areas (dRM = 0.373), although in this case, the difference is statistically significant.

Finally, the impact of the URBAN programme on the improvement of human capital in the neighbourhoods is quite evident: 70% of the experimental areas achieve a greater improvement than the control areas (dRM = 0.510). This could mean that, at least in part, some improvements depend on generational renewal. Young residents have brought about this change because they have reached the level of university education or entered the employment market in occupations better than those usually found among the neighbourhood’s employed population in 1991 due to their better educational level (as regards socioeconomic status). However, this generational thesis should be explored in more detail.

Overall, the results would indicate that, although moderate, the programme has achieved some socioeconomic revitalisation in the neighbourhoods, thus, in accordance with its policy frame (see Chapters 1 and 6). The improvement in economic activity in the experimental neighbourhoods has been higher than in the control areas, but it seems that the rate of employment has also improved to a greater extent, albeit very moderately, as has the socioeconomic status of residents and their level of education, in a slightly more straightforward way.

This is highlighted by the fact that the models that analyse only stayers show more evidence of the impact than the models that also include incomers (dRM values in Fig. 8.3), although the magnitude of the change is greater when the latter are also included (Table 8.1). New residents might have a higher socioeconomic status or level of education than the residents in 1991, but the neighbourhoods in which the programme intervenes do not appear to show a superior pattern of change to the control neighbourhoods when new residents are included in analyses together with stayers. Therefore, there do not appear to have been clear gentrification patterns in experimental neighbourhoods promoted by URBAN, but instead, patterns that point to their socioeconomic revitalisation. Our post-intervention measurement is very close to the completion date of the projects. This might not show the change assumed in the gentrification thesis, which might need more time to appear. However, evidence show projects promoted changes pursued by the EU urban initiative among stable residents.

Place-Based Integrated Strategies and the Puzzle of Residential Mobility: on Gentrification, Revitalisation, and Socio-Spatial Inequalities Reproduction

In this chapter, we have sought to provide some arguments and strategies to improve the analysis of the impact of the integrated strategy, using the most commonly available data to approach this type of analysis (aggregated data at the sub-municipal scale). To this aim, we have indicated the importance of considering exposure to the programme in more detail rather than assuming that the integrated strategy only involves a change in the neighbourhood as a structure of opportunities; that is to say, that their impact is derived only from the effects of ‘contextual exposure’ to them. Based on this assumption, we have pointed out the need to consider the phenomenon of residential mobility, as it relates to assumptions about policy contextual exposure and, therefore, to the adequacy of the programme’s underlying theory.

In response to these two issues and the data available, we have performed controlled comparisons according to programme selection bias (experimental areas vs. control areas) and programme exposure derived from residential mobility (all residents vs. stable residents). By establishing these two ‘controls’ in our comparisons, derived from analytical arguments, we have been able to show more clearly the impacts of the URBAN programme. Very briefly, the results of applying this strategy show that the projects improved the economic activity of the urban areas where they were applied, at least in their business density (as we have been able to measure this aspect here). But also that, although moderate, they brought about improvements among traditional residents, more exposed to their actions because they remained in the neighbourhoods from the beginning to the end of the programme.

More generally, the strategy used and its results could be useful when discussing the effect of such initiatives in terms of the revitalisation or gentrification of the urban areas where they are applied. When considering all residents, the change observed in neighbourhoods is more intense than when considering only stable residents, and policy impacts do not exist. However, when considering only the second group, differences are low but the programme’s impacts become more evident (differences in the improvement patterns of experimental and equivalent areas). Does the former imply evidence of gentrification processes because the change is more intense? Are new residents with a higher socioeconomic status replacing traditional residents with a lower socioeconomic status? Even if the answer to the second question is yes, the arrival of these new residents or incomers in the neighbourhoods does not seem to create much greater change in the socioeconomic status of experimental areas than in the control areas.

Does the impact revealed by analyses of stayers provide evidence of neighbourhood revitalisation? In this regard, we can at least indicate that the improvement observed among these residents is slightly higher in areas where projects are developed. The evidence points more to processes of revitalisation rather than gentrification, although to be more conclusive, we would need to continue using the proposed strategy, incorporating other indicators or seeking to evidence the impacts of the projects over a longer term.

Moreover, the analytical ideas and analyses show the importance of paying attention to the residential mobility phenomena to know better the effects promoted by the urban integrated strategy in neighbourhoods. This phenomenon informs about the exposure to the programme and, therefore, its potential impact. Therefore, residential mobility could be a cause explaining the effects promoted by the integrated strategy, or at least a factor explaining its capacity to improve residents’ quality of life. However, this implies residential mobility could also be a consequence (an effect) of these policies that could hide their impacts if residents who improve their situation move to other neighbourhoods (in the same city or other cities in the metropolitan area). In sum, residential mobility suppose a puzzle for area-based integrated strategies: it could be a factor explaining the usually low impacts found on previous evaluative exercises, meaning the design of these urban initiatives should pay attention to this phenomenon by including policy actions specifically oriented to promote and encourage residential stability in targeted territories. Gentrification could be a risk to these area-based initiatives as replacement of traditional residents by others with higher socioeconomic status. However, the ‘upgrading’ residential mobility that could promote these initiatives also suppose a risk regarding socio-spatial inequalities reproduction in specific urban areas due to the ‘runaway’ of those who improve their socioeconomic conditions thanks to the public intervention.

Like other possible ones, the proposed strategy to analyse this issue is not without limitations. It only allows for an approach that considers contextual exposure to the programme, without being able to provide evidence on the effect of targeted exposure (e.g. motivational mechanisms targeted at specific groups of residents). Even so, if the hypothesis that the projects produce upward residential mobility is true, or because the effect of specific actions requires a broader timeframe to be evidenced in terms of improvements among residents, we have adopted a conservative strategy in analysing the impact of the programme. In any case, the strategy put forward here should be taken as a proposal. Using the most commonly available ecological data sources (aggregations at the urban area level, for instance, census data) offers an approach to analysing the impact of integrated urban development initiatives. The strategy is based on analytical ideas about the theory of this type of public policy (contextual mechanisms and exposure) and other phenomena related to urban change (residential mobility) that must not be excluded from any evaluation or assessment of urban integrated strategies promoted by the EU or other similar policies adopting a multi-level policy mixes character (see Chapter 1).

Annex

Economic activity in the neighbourhood 1991–2001 (Mean [standard deviation])

  

Total 1991

Total 2001

Business density

Control

7,273

24,416

  

(8,65)

(23,665)

 

Experimental

8,300

30,687

  

(8,566)

(25,231)

 

Total

7,461

25,566

  

(8,608)

(23,974)

  1. Experimental areas = 22; Control areas = 98
  2. Source E-Informa

Socioeconomic conditions of neighbourhood residents 1991–2001 (Mean [standard deviation])

  

Total residents in 1991

and 2001

Stayers:

stable residents in 2001

  

1991

2001

2001

Employment rate

Control

80,096

84,274

83,025

  

(6,877)

(5,856)

(6,151)

 

Experimental

79,226

83,824

82,885

  

(4,489)

(4,423)

(4,83)

 

Total

79,936

84,191

82,999

  

(6,498)

(5,607)

(5,913)

Socioeconomic status

Control

0,497

0,529

0,509

  

(0,121)

(0,114)

(0,115)

 

Experimental

0,501

0,537

0,522

  

(0,108)

(0,118)

(0,113)

 

Total

0,498

0,531

0,512

  

(0,118)

(0,114)

(0,115)

Universitarys

Control

10,881

18,562

14,013

  

(9,148)

(11,956)

(10,208)

 

Experimental

10,554

18,703

14,488

  

(6,543)

(9,347)

(7,532)

 

Total

10,821

18,588

14,100

  

(8,705)

(11,487)

(9,746)

  1. Áreas experimentales = 26; Áreas de control = 98
  2. Source Spanish Population and Housing Census (INE)