Are Coworkers in the Italian Peripheral Areas Performing Better? A Counterfactual Analysis

Abstract

Coworking spaces are “serendipity accelerators” designed to host creative people and entrepreneurs. While recent literature has started exploring the indirect effects of coworking spaces on the local context, little is still known on how coworking spaces may directly affect the coworkers’ economic performance and wellbeing. Using a novel dataset based on a survey of 326 CWs working in the Italian coworking spaces in 2018, this paper explores the potential economic impact for coworkers, depending on whether a coworking space is localized in a peripheral or an urban area. Through a propensity-score matching approach, we found that being located in a peripheral area for coworkers may represent an opportunity to earn more than working in an urban center. The same holds for the organization coworkers belong to.

1. Introduction

Coworking spaces (CSs) are regarded as “serendipity accelerators” designed to host creative people and entrepreneurs who endeavor to break isolation and to find a convivial environment favoring the development of collaborative communities [1,2], whose social, institutional, cognitive, and organizational proximities [3] may lead to an increase in informal exchange, collaboration and interaction with others, knowledge transfer, and business opportunities [1]. So far, CS has proven that it could be a profitable business and expand the scope of companies by adding supplementary services welcomed by entrepreneurs and the self-employed [4]. CSs host different professionals, specifically those engaged in start-ups, early-stage entrepreneurs, freelance, self-employed, and independent knowledge workers, who need social and professional interaction to overcome the risk of isolation and loneliness [1].

Although CS is mainly an urban phenomenon, the last few years have experienced a flouring of CSs in peripheral areas. This trend seems to have increased during the COVID-19 pandemic, where specific policies have promoted the growth of CSs to host knowledge and remote workers [5].

The potential benefits deriving from a workplace located in a remote area concern a higher general wellbeing, lower congestion, less polluted air, lower cost location, exploiting institutional leeway, etc. [6]. The literature exploring the effects of CSs is still scant. It focuses on the “direct” effects on the individuals (the coworkers, CWs) and the “indirect” effects on the local context [7]. Within this context, the present paper aims to fill the gap in the literature investigating the potential economic effects for the CWs of working in a CS in a peripheral area instead of an urban area. Data come from a survey of coworkers in Italy in 2018; 326 coworkers, homogeneously distributed in the country, answered the online questionnaire. The survey was composed of socioeconomic questions, e.g., to reveal if their economic incomes and the incomes of the organizations they belong to have found significant advantages from the localization in CSs, disentangling between core and peripheral areas.

We assume that the economic benefits deriving from localization in a peripheral area favor the CWs, and their belonging organizations.

Other significant questions are used to build a set of control variables (x_i) in which, other than sociodemographic controls (gender, education, creative intensity, distance from the CS), three types of proximity à la Boschma (social, institutional, and organizational) and a series of observed benefits (professional, social/friendly, formative/educational, instrumental, space) by the CWs, are considered.

The paper is composed of five sections. The introduction is followed by a literature review about CSs in peripheral areas and the direct and indirect effects. Section 3 is dedicated to data and methodology; it also presents the main descriptive statistics and the balancing tests to determine the validation of the proposed model. The results of the empirical analysis are presented in Section 4; Section 5 discusses the results and concludes the paper by putting forward new research avenues.

2. Coworking Spaces and Their Effects

The CS is defined as “third place” [8], where the users (CWs) can interact and develop a “community”. The CWs are freelancers, self-employed individuals, entrepreneurs, dependant contractors, consultants, and employees with diverse professional profiles and competencies. Their fields range from the creative industry: from architects, designers, journalists, etc., to engineering and digital sectors, software developers, consultants, etc. [1,9,10]. The CS offers them a workplace and several facilities, e.g., secretary services, printing, Wi-Fi, shared rooms (kitchen, garden, etc.), meeting rooms, and training courses to increase business opportunities. The proximity measures a là Boschma [3]—geographical, social, institutional, cognitive, and organizational—represent the spirit of coworking, where interaction and knowledge exchange are pivotal for such working communities [11]. The coworking phenomenon was officially born in San Francisco (US) in 2005.

The coworking spaces are mainly an urban phenomenon [12] because the CWs are primarily knowledge workers and creatives willing to live in cities to exploit urban amenities [13]. This explains why the literature on new working spaces, specifically CSs, is mainly concerned with large urban areas and metropolitan regions [12]. Nevertheless, there has been a growing awareness and interest in the potential of small cities, peripheral, peri-urban, and rural areas to attract CSs [13]. Specifically, the COVID-19 pandemic has brought attention to more peripheral and rural working environments, where CSs might host remote workers, offering them a workplace away from congested urban environments, with obvious lower risks of contagion and more advantageous living conditions in terms of costs and commuting reduction.

As underlined by the literature, CSs in small and medium-sized cities and peripheral areas can be considered tools for regeneration purposes, place marketing, and attracting economically active individuals and their families. Felton et al. [14] stated that the dense proximity cluster networks of the inner city are not the only environment in which creative industries operate because the geography of creative industries is more complex than simple concentric circle models suppose. Indeed, the simple co-location itself may not necessarily lead to networking, interaction, and collaboration and thereby to knowledge creation. At the same time, community facilitators, such as CS managers, may play an essential role in enabling more synergies to stimulate encounters and partnerships inside the trust-based community-oriented environments [15].

Therefore, the sense of community can be considered a solution offered by CS to overcome the issue of social isolation also in less central areas. Recently, medium-size companies have also tended to locate their business in CS.

According to Jamal [16], a CS in a mid-sized city downtown promotes urban renewal, preserving affordable space for new enterprises in rapidly gentrifying areas. An example of CSs in peripheral areas is provided by Fuzi [15] in the analysis taking place in South Wales, showing that CSs can support entrepreneurship in sparse regions. Moreover, coworking spaces in rural and peripheral regions are recognized as drivers for social cohesion and economic development [17]. The rural areas in Catalonia [18], where technological and digital advancements have pushed workers from urban centers to rural areas represent an interesting case.

Recent literature is focusing on the effects of CSs and, more generally, collaborative and flexible spaces, disentangling between direct effects (on the individuals) and indirect effects (on the local environment) (

Table 1. Effects typology. The positive direct and indirect effects of CSs ¹.

Direct effects	Coworkers
	Cost savings (office rental, office energy consumption, employees’ commute times) Reduce risks of isolation Increase meeting opportunities Boost business collaboration and promote innovation Foster employee work productivity and working efficiency Reduce CW commute times and earnings growth Boost CWs job satisfaction and wellbeing
Indirect effects	Urban spaces
	Confirmation of urban attractiveness Development of spontaneous aggregation in districts Episodic transformation in the public space (temporary installations, permanent/new equipment)
	Practices
	Contribution to the development of innovative services Extension of daily and weekly cycles of use Episodic participation in the strengthening of communities (i.e., Social Streets) Revitalization of existing retail and commercial activities Strengthening mini clusters of creative and cultural productions
	Environment/Planning
	Reduction of pollution Traffic congestion Energy use patterns both at home and associated with travel/transportation

¹ Source: adapted by Mariotti et al. [7].

).

The first concerns how working in a CS by a CW impact his/her economic performance and wellbeing. As stated by Mariotti et al. [7] (pp. 175–176), it can do so by: cost savings (e.g., office rental, office energy consumption, and employees’ commute times) [19,20]; reducing risks of isolation; increasing meeting opportunities, boosting business collaboration, and promote innovation [21,22]; fostering employee work productivity [23], working efficiency, economic performance/earnings growth [6]; boosting coworkers’ job satisfaction and wellbeing [24,25]; (for a review see also [26]).

As found by a few studies, the positive effects of working in a CS are related to the proximity measures a là Boschma [3]—geographical, social, institutional, cognitive, and organizational proximities—CWs can experience in a CS. These proximities, indeed, may foster informal exchange, collaboration, and interaction with others, knowledge transfer, and business opportunities within the CS [1,11,27]. Similarly, Parrino [28] found that social and organizational proximity promotes collaboration and knowledge exchange among CWs.

No specific study has focused on the impact on CWs distinguishing between those working in core and peripheral areas. The present study aims to fill the gap in the literature by exploring whether and how working in a CS, located in a peripheral area instead of an urban area, affects the economic performance of CWs and the companies belonging to.

To reach this goal, the following two main hypotheses were tested:

Hypothesis 1.

Being in a peripheral area fosters higher incomes for the CWs.

Hypothesis 2.

Being in peripheral area foster higher incomes for the CWs’ organization.

In addition to direct effects, CSs may produce indirect effects on their local context. CSs may become drivers of economic change while retaining the creative class and knowledge workers in the periphery, thereby increasing the competitiveness and performance of remote areas [15,29] and improving the urban quality and individuals’ wellbeing [30].

The literature exploring the effects of CSs on the local context is scant. The few studies analyzing the indirect effects of CSs on the local environment have focused on the built space, environment, organizational/working practices, urban planning, and policy design [7]. Yu et al. [20] have carried out a literature review on the impact of future flexible working model evolution on the urban environment, economy, and planning (Table 1). The growth of CSs is considered an opportunity to foster socioeconomic development and urban regeneration [31]. Moreover, CSs can impact urban planning since they lead to less traffic congestion, promotes a collaborative culture, and spread workers towards regional areas, bringing about opportunities for urban infrastructure planning [20,32].

Mariotti, Akhavan, and Di Matteo [7] explored the indirect effects of CSs on the local context, disentangling the CSs located in Italian pole municipalities and in non-pole municipalities. Their analysis showed that, on average, CWs in non-pole municipalities, compared to those in pole cities: (i) perceived a higher positive impact of the CS in the urban context; (ii) have a lower educational level; (iii) tend to work in a creative sector; (iv) declared to be more satisfied; (v) live closer to the CS; (vi) experienced higher social and organizational proximity, and lower institutional proximity; (vii) have created new professional relationships, and had the chance to access new information channels and new training opportunities inside the CS. These results were corroborated by a counterfactual analysis, showing that non-pole areas experienced a higher and more positive impact on the local environment than those located in a pole municipality.

3. Research Design

3.1. Methods

This study proposes propensity-score matching (PSM) as a counterfactual method based on the selection of observables [33]. PSM does not require linear relationships between the potential outcome (y) and the set control variables (x_i), and it is a non-parametric estimator to evaluate the treatment parameters (ATEs). The propensity-score (PS) is the conditioned probability to receive a certain treatment (w) given the pre-treatment set of variables, and it is useful to reduce the observed bias in selection models [34] (p. 288), [35] (p. 310). Since the PS is unknown, it needs to be estimated through a probability model (a probit model in this case), as follows:

p(x_i) = Pr{w = 1|x_i} = Λ[f(x_i)]

(1)

where f(x_i) is a linear function of the selected covariates. The PS need to satisfy two main properties: (i) the balancing of the pre-treatment variables where given the p(x_i),

w⊥x_i|p(x_i)

(2)

the treatment (w) and the covariates (x_i) are independent; (ii) the conditional independence assumption which, given the p(x_i):

(y₁;y₀)⊥w|p(x_i)

(3)

meaning that, once the knowledge of the factors affecting the sample selection is taken into account (or controlled for), the condition of randomization is restored [33] (p. 19). Thus, the general formula to estimate the average treatment effect on the treated units (ATET) is:

ATET = 1/N_T ∑ _{(i ∈ {w = 1})}[y_1i–∑_{(j ∈ C(i))}h_ijy_0j]

(4)

where N_T is the number of treated units; i is the treated unit with the characteristics assumed by the x_i; j is the non-treated unit matched with the i-treated unit based on the PS; h_ij are the weights (ranging between 0 to 1) to be applied for each j matched to i; the set of C(i) individuates the i- neighborhood for each j matched to i; w is the treatment status and identifies the threshold between the treated and the control group [33] (p. 84).

The following two types of matching will estimate the effect of the treatment status: (i) the radius matching, which is a variant of the nearest neighbor that takes into account the bad distances between pairs of scores through the introduction of a pre-determined tolerance level [36] (pp. 160–161), the caliper, beyond which too fair distant matches are eliminated, and the C(i) of the (4) are defined as:

{j:‖p_i − p_j‖ < r}

(5)

where r is the caliper, which is set here at 0.1 as default; (ii) the kernel matching, in which every treated unit is matched with a weighted average of all control units with weights that are inversely proportional to the distance between the treated and the control units [33] (p. 87), where, given the (4), h_ij has to be substituted with:

h_ij = K((p_j − p_i)/b)/∑_{(K ∈ [w = 0])}K((p_k − p_i)/b)

(6)

where K is the kernel function, and b indicates the bandwidth parameter (0.6 in this case), while the standard errors are bootstrapped (50 repetitions).

3.2. Data

The variables used in this study originate from an online survey addressed to the managers of all CSs located in Italy; 326 CWs working in 138 CSs in Italy in 2018 (25% of the total CSs) answered the survey. The survey mainly included socioeconomic questions, and CWs were asked to reveal if their economic incomes and the organizations’ they belong to have found significant advantages from their work localization in a CS. These questions will constitute the two outcome variables (y₁, y₂) to be tested with the counterfactual model, where y₁ expresses the individual incomes of the CW while y₂ indicates the organization incomes.

At the beginning of January 2018, there were 548 CSs in Italy. Following the definitory criteria introduced by the SNAI in 2014 (Strategia Nazionale per le Aree Interne—National Strategy for Inner Areas) [37,38], we noticed that about 80% of CSs were located in “single municipality service centers” (poles) and “multi-municipality service centers” (intermunicipal poles), 16% in “belt” peri-urban municipalities, while only the 3.5% were localized within “intermediate”, “peripheral”, and “ultra-peripheral” areas (Figure 1).

The SNAI categorization refers to the prior identification of specific essential dimensions that are mandatory for a municipality to be considered a “service center” area, including (1) school—the presence of at least an upper secondary school; (2) health—the presence of at least one hospital offering the first level DEA (department for urgencies and emergencies); (3) mobility—delineated by the minimum presence of a silver grade railway station. Specifically, “service centers” are those municipalities that simultaneously own the three abovementioned dimensions, or at least one out of the three, and they are encompassed within the broader category of poles [38]. On the other side, those municipalities without such dimensions are later classified on a travel-time distance criterium from the “service centers”, and three types of areas are identified, i.e., intermediate, remote, and ultra-remote. Such areas are encompassed within the broader category of “peripheral areas,” that are “areas at some considerable distance from hubs providing essential services (education, health, and mobility), with a wealth of key environmental and cultural resources of many different kinds, which have been subject to anthropization for centuries” [37].

To smooth the research framework, in this paper, we labeled all the “service center” municipalities as “urban areas” and all the other municipalities as “peripheral areas”. This allowed easy comparison between CWs working in substantial heterogeneous areas. In our context, the ideal treatment status (w) concerns the geographical localization of CSs and therefore the workplace of CWs in a “peripheral area”: it takes the form of a dummy variable assuming the value 1 if the CWs are localized within a “peripheral area”, while it takes the value 0 if the CWs are based outside the peripheral areas, namely within an “urban area”. This treatment assignment does not correspond to a specific program or a policy choice. Still, it simply reflects the difference in the economic benefits that CWs may receive, whether their location is inside or outside a “peripheral area”.

3.3. Descriptive Statistics

On such premises, we recall the H₁ and H₂ hypotheses to be tested in Equation (4), where we assume that the economic benefits deriving from localization in a “peripheral area” favour the CWs and their organizations.

Other significant questions of the survey will constitute the set of covariates (x_i) allowing to control for sociodemographic characteristics (gender, education, creative intensity, distance from the CS), three types of proximity à la Boschma (social, institutional, and organizational), and a series of observed benefits (professional, social/friendly, formative/educational, instrumental, and space).

When looking at the simple comparison of the mean of the selected variables (

Table 2. Mean comparison of the variables between “treated” and “control” groups; standard errors in parentheses.

Variables	“Treated” Units (CWs in Peripheral Areas)	“Control” Units (CWs in Urban Areas)	Range
Potential outcome
Coworkers’ incomes (1 = increases perceived)	0.705 (0.469)	0.385 (0.487)	0–1
Organisations’ incomes (1 = increases perceived)	0.647 (0.492)	0.281 (0.450)	0–1
Demographics
Gender (1 = male)	0.647 (0.492)	0.559 (0.497)	0–1
High-skilled (1 = masters’ degree or superior)	0.588 (0.507)	0.595 (0.491)	0–1
Creative intensity (1 = ‘creative’ type of job)	0.823 (0.392)	0.731 (0.443)	0–1
Distance home/coworking	1.764 (1.032)	1.893 (0.999)	1–4
Type of proximity
Social	1.941 (0.242)	1.653 (0.608)	0–2
Institutional	0.764 (0.903)	0.705 (0.837)	0–2
Organisational	1.764 (0.664)	1.546 (0.810)	0–2
Observed benefits
Professional	2.588 (0.795)	2.197 (0.981)	0–3
Social/friendly	2.235 (0.831)	2.203 (0.953)	0–3
Formative/educational	2.352 (0.996)	1.644 (1.157)	0–3
Instrumental	2.176 (1.131)	2.090 (1.002)	0–3
Space	2.764 (0.437)	2.521 (0.662)	0–3
Observations	17	309

) between CWs localized in peripheries (treated) and CWs localized in urban areas (control), we observed that in most cases, the values for treated subjects are relatively higher than the control subjects. A counterfactual model will empirically test the hypothesis that being located in a peripheral area—both for CWs and their belonged organizations—is more performing than being in an urban area.

The set of control variables (x_i) thus permits selecting those individuals among the two groups (treated and controls) that are very close in terms of characteristics, but that differ only based on the treatment assignment (w), which corresponds in this case to the localization in a peripheral destination. The PS will help restrict the field only to those observations in the common support. In other words, based on the control variables, the PS acts as a selector of those individuals in the two very similar groups, but that differ only due to their geographical localization inside or outside a “peripheral area”. Once the structural differences are removed, the counterfactual design will tell us if the localization of CWs in a “peripheral area” (instead of “urban”) might determine an increase in their economic performances.

3.4. Balancing Tests

We verified matching quality before moving to the results by looking at the PS density before and after the matching procedure. The PS was built upon a probit regression, where the treatment status (w) was set as the dependent variable of the set of control variables (x_i). The coefficients of x_i should not be statistically significant against the w, meaning that the set of control variables should not influence the probability to fall inside or outside the treated or the control group. This was the first measure of goodness of fit of the research scenario, facing the conditional independence assumption.

Then we computed the balancing properties for each block individuated by the procedure, verifying that the balancing properties test was satisfied. In this case, we had a final number of blocks equal to 5, and the region of the common support was (0.006, 0.264). We do not report the complete tests for each block (that are available upon request), but we show the graphical overlapping of the PS before and after the matching (Figure 2). Visually, this was enough balanced to guarantee the robustness of the proposed counterfactual analysis.

4. Results

Table 3. The potential impact of peripheral localization on CWs’ incomes ².

	Radius Matching	Kernel Matching
Outcome: y1
Treatment: w
ATET	0.253 ** (0.119)	0.223 * (0.121)
N° of treated units	17	17
N° of control units	235	235

² Note: standard errors in parentheses and bootstrapped at 50 repetitions in kernel matching; ** indicates significance at p < 0.05; * indicates significance at p < 0.10.

shows the potential effect of localization in a sparsely populated “peripheral area” on the CWs’ economic performances, where the counterfactual group is composed of subjects operating in densely populated “urban areas”. Both the estimations were positive and significant, meaning that being in a peripheral area for CWs may represent an opportunity to earn more than working in an urban center. It was observed a positive deviation from the mean of the dependent variable (y₁), which was around 0.2. This led us to assume that peripheral areas are more economically profitable for CWs.

Table 4. The potential impact of peripheral localization on CWs’ organizations incomes ³.

	Radius Matching	Kernel Matching
Outcome: y2
Treatment: w
ATET	0.324 *** (0.123)	0.319 *** (0.122)
N° of treated units	17	17
N° of control units	235	235

³ Note: standard errors in parentheses and bootstrapped at 50 repetitions in kernel matching; *** indicates significance at p < 0.01.

shows the potential effect of localization in a peripheral area for the organizations’ incomes to which the CWs belong. As previously, a positive and entirely significant impact is observed for the ‘treated’ individuals. In other words, if an organization collocates its employees in a CS settled in a peripheral area, the organizations’ incomes would probably be more performative as against the case in which the employees had been placed in a CS located in an urban area.

The coefficients related to the organization’s incomes show even a higher value (and better significance) if compared to the CWs’ incomes, which is around 0.3 points above the mean of the dependent variable (y₂). This means that localization in peripheral areas might be fruitfully profitable both for the coworking managers and for the coworking users. As per the y₁, also in the case of y₂ we derive that localization in peripheral areas is more profitable for organizations than localization in urban areas.

A further robustness check has been made by computing the ATET through the bias-corrected standard errors [39], and the results generally hold. Moreover, it is also visible that the number of control units in Table 3 and Table 4 are less than those indicated in the full sample because only individuals in the “control” group that are perfectly comparable with those of the “treated” group are considered by the PSM within the common support.

5. Conclusions

This paper has shown that CWs working in CSs settled in peripheral areas may have higher chances to increase their revenues than CWs working in CSs in urban areas. The same holds for the companies the CWs belong to.

This empirical evidence might be helpful for the future planning strategies that CSs and CWs will implement. Moreover, these results might foster relevant policy implications, especially during and after the COVID-19 pandemic, thus fostering the growth of CSs in peripheral areas. These areas, probably because of their relative geographical isolation, have suffered less from the spread of COVID-19. They have been the first to remove the ‘social distancing’ and restart an everyday life, thus being considered ‘safer places’. On the other hand, the (full or partial) lockdown of activities imposed by a large number of cities and metropolitan areas from all over the world represented a possible threat for the rise of CWs creative and innovative classes and the CSs themselves.

The development of CSs in peripheral areas might also be a positive alternative to host remote workers, mainly specialized in knowledge-intensive services, who started working from home to guarantee social distancing.

Nevertheless, there is an urgent need to enlarge the (CS) toolkit for local development in smaller towns and rural areas [26]. Indeed, the COVID-19 pandemic has affected working typologies, office premises, and work geography, making suburban and peripheral areas more attractive than before. As stated by Manzini Ceinar et al. [40], the following recurring trends can be recognized: (i) the willingness of companies to downsize or “de-densify” their offices by relocating employees in other locations (hubs) different from the main headquarter and promoting remote working; (ii) the tendency by freelancers and digital nomads to move to suburban and peripheral areas to experience a higher quality of life, stimulating suburbs economies and catalyze phenomena such as what has been defined in Italy “South-Working” [41,42]; (iii) new working spaces such as CSs are changing their business model to be more attractive for teleworkers [43,44]; (iv) local authorities are using public spaces and public services, such as public schools and libraries, to relocate employees and students (such as the Scuola Diffusa diffused schooling experiment launched in Reggio Emilia in 2020).

From the perspective of local administrators of peripheral municipalities, attracting CSs located in urban poles that have been greatly affected by the COVID-19 might be a good strategy, maybe offering them incentives and/or tax benefits to relocating to their municipality. This suggestion may also be of great importance both for CSs managers and CWs to reschedule the resumption of their activities. Suppose we consider that being in such places may be profitable both for the hosting destinations and individuals. In that case, the relocation hypotheses of CSs and CWs from the urban poles to remote areas should not be underestimated.

Nevertheless, the scenario we depicted presents some criticalities and uncertainties. First, creatively innovative and skilled workers are more willing to live in metropolitan areas [45]. Suppose the locus amoenus patterns of remote areas improve the quality of life. In that case, the lack of infrastructures (i.e., broadband and low transport accessibility) can inhibit every type of work, and suppose the policymakers promote the location of CSs in peripheral areas by hosting them, for instance, in public libraries or other public premises [46]; in that case, it is necessary to verify: (i) the potential demand of CWs and CSs managers and their willingness to pay for these services, also evaluating the sustainability of the project in the long run; (ii) the technical feasibility and start-up costs; (iii) the risk to denature the coworking concept, due to the loss of dynamism and involvement in sharing the space in favor of a more static and practical use of it. Beyond the complexity of the time being, this scenario is characterized by the uncertainty of the operation’s time lane and, therefore, by the impossibility of estimating the investment’s payback time.

The debate about the possible strategies to pursue by the CSs and the CWs in the post-emergency phases due to COVID-19 has just begun. It will require time and dialogues among coworking managers and public subjects to get a way forward, where place-sensitive policies are needed to get effective results in peripheral areas [47,48]. A growing shared sentiment tends to rethink and reconsider the possible key role of the peripheries in the period of socio-economic ‘reconstruction’ post COVID-19.