Abstract
There is currently considerable interest in workers performing tasks from a variety of workplaces, such as co-working spaces, transport-networks and cafés. However, it remains difficult to ascertain the extent to which this workplace mobility is altering urban economic geography, since most analyses of the location of economic activity in cities are based upon census-type data that assume a unique place of work for each worker. In this paper I propose a framework that extends the concept of place of work: work is probabilistically assigned to different types of workplace according to the proportion of work time spent in each. The limitations of census data are discussed and illustrated, after which the framework is operationalised in an exploratory survey. Census data suggest a modest increase in workplace mobility, with most work still taking place either at home or in a fixed workplace. The paper’s principal contribution is to explain these data’s limitations and show how work location can be operationalised as a probability space.
Introduction
There appears to be a revolution in the way economic activity deploys within cities: co-working spaces (Moriset, 2014), trains (Axtell et al., 2008), cafés (Liegl, 2014), cars (Hislop, 2013) are all now locations from which work tasks are performed. Media reports reveal that co-working companies such as Spaces and WeWork are leasing prime space in central New York (Mashayekhi, 2018), whilst elsewhere freelance workers have self-organised into systems such as Hoffice, where they co-work in each others’ living rooms (Mok, 2016). Major corporations have taken note: they are reorganising their workspaces (and work practices) as ‘activity based work meets millennials’ demands and expectations in the workplace’ (Morley, 2018) – this entails allowing employees flexibility to work in a variety of spaces within the office but also to perform work tasks in other locations if this is appropriate.
Notwithstanding these changes, most spatial analyses of the urban economy continue to rest on the assumption that assigning workers to a single place of work (usually their establishment of employment or their home) provides a good approximation of the geography of urban economic activity. Chicago sociologists of the 1920s, who analysed places of residence, places of work and commutes (Park et al., 1925), as well as current researchers who assess how knowledge workers cluster and live relative to their place of work (Duvivier et al., 2018; Frenkel et al., 2013; Shearmur, 2012), make this approximation. This is a heuristic device that allows for tractable analyses and that is compatible with available data: in the light of current changes, this paper explores whether the approximation remains justified and suggests a complementary way of conceptualising the location of economic activity.
Of course, all work – understood in this paper as remunerated work – necessarily occurs at a location: but workers’ tasks seem decreasingly to be occurring in one place (Hermelin and Trigg, 2012; Shearmur, 2018). Already in the 1960s Hägerstrand had developed time-geography, that is the idea that individuals’ daily activities occur along a spatial path, punctuated by fixed locations and times – such as picking up children from school, sleeping at home and being at work (Pred, 1977): time-geography recognises the time/space component of daily life, but reflects the common assumption that work occurs at a specific location. The idea that movement is critical to understanding social activities and structures was reinvigorated in the early 2000s by Urry’s sociology of mobilities (Urry, 2000, 2007): without direct reference to Hägerstrand, 1 Urry makes a general argument for the importance of breaking with traditional approaches to sociology, which focus on cross-sections and on specific places without embracing movement (whether geographic or across social boundaries). Contemporaneously, Bauman (2000) made similar points, speaking of ‘liquid modernity’, characterised by the impermanence of social structures and location.
Whilst Bauman’s and Urry’s ideas have spawned a growing ‘mobilities’ literature, epitomised by the eponymous journal, it remains unclear whether analysis of the urban economy requires new methods and concepts in the light of workplace mobility. Indeed, urban economic geographers have developed robust theories, backed by evidence, about where establishments choose to locate and, by extension, about where (most) people work (e.g. Dicken and Lloyd, 1990; McCann, 2013). Geographic distributions of employment by place of work can be successfully interpreted in the light of these theories, notwithstanding the discomfort that some analysts have expressed: Coffey and Shearmur (2006), for instance, acknowledge a ‘new generation of mobile workers: those who work from a combination of locations, including home, cars/trains/planes, customers’ premises, and the central office of the company’ (2006: 251), yet they analyse place of work because ‘few reliable statistics exist concerning the incidence or rate of growth of this phenomenon’ (2006: 251). On the one hand, therefore, traditional methods and analyses of place of work continue to provide results that are coherent with theory and that have practical relevance (they reveal and describe phenomena that can be observed, for example Duvivier et al., 2018): yet, on the other hand, the nature of work location has been evolving since the early 2000s and this evolution seems to be accelerating.
This paper has two aims. The first is to suggest a framework for operationalising place of work that does not require the simplifying assumption of assigning each worker to a single location. The second, empirical, is to use primary data derived from an exploratory survey to operationalise this alternative framework. Secondary census data covering Canada’s ten largest urban agglomerations are used to frame the survey and to illustrate the limitations of current data.
The paper proceeds as follows. In the next section a brief overview of the recent literature on employment and mobility, as it pertains to the location of economic activity, is presented. This is followed by a discussion of how work can be broken down into tasks and of how intra-metropolitan work location can be conceptualised as a probability space. Data and methods are then discussed, with emphasis on the limits and possibilities of the survey: the survey data are exploratory and are used to illustrate how the concepts and framework introduced in this paper can be operationalised.
We conclude that the new framework can indeed be operationalised, in part because most workers perform work tasks from a finite number of locations, predominantly (although not exclusively) from a usual place of work or from home. Although work mobility is now greatly facilitated by information and communication technologies (ICT), mobile communications in particular, our exploratory survey and secondary data analyses corroborate other recent studies (Felstead, 2012; Ojala and Pyöriä, 2018; Vilhelmson and Thulin, 2016), only revealing marginal changes (so far) to the location of work. This does not preclude the possibility of more marked changes for specific types of work or in certain neighbourhoods.
Mobilities, work tasks and work location
Mobility and the location of work
Mobility has, since the early 2000s, become a key lens through which social processes are studied (Bauman, 2000; Rainie and Wellman, 2012; Urry, 2000, 2007). Likewise, as the workplace has been reconfigured under the influence of communication technologies (Hermelin and Trigg, 2012) and changing management practices (Boltanski and Chiapello, 1999; Friedman, 2014), so questions are increasingly being asked about the nature of the workplace (Valenduc, 2019). However, these questions tend to be asked either from the perspective of organisations (and of the employment relationship) or of particular work practices such as those observed in co-working spaces (Gandini, 2015; Merkel, 2019), trains (Axtell et al., 2008) or cars (Hislop, 2013). Some of this research seeks to re-think the relationship between work and (diverse) workspaces (Hislop and Axtell, 2009); some of it examines how workers cope with the tensions caused by implicit or explicit management control in spaces that were previously exempt (Sewell and Taskin, 2015); some of it studies the complexities of balancing work and family life (Schieman and Young, 2010). Other research highlights the injustices that mobility and ICT can impose in certain contexts, such as zero-hour contracts (Sheller, 2018).
Although urban geographers are aware of work mobility (Coffey and Shearmur, 2006), in their analyses of the urban space economy they tend to adopt the simplifying assumption that each worker can be spatially assigned to a single place of work (POW) (e.g. Coll-Martínez, 2019; Duvivier et al., 2018; Folmer and Kloosterman, 2017). This research reveals the changing location of jobs and is corroborated by qualitative work on cities (Shearmur and Hutton, 2011) and by location theory (McCann, 2013): such empirical and theoretical validation suggests that the geography of economic activity derived from traditional POW analysis is empirically robust. However, as Hermelin and Trigg (2012) argue, it may be increasingly important to distinguish between where employment is officially assigned (that is, the address of the employer) and where work tasks are actually performed. As workers are able to perform work tasks in transport, in parks, at home, in co-working spaces, whilst waiting for a doctor’s appointment and so on, new insights may emerge if these multiple work locations can be incorporated into urban economic analysis (Shearmur, 2018).
Notwithstanding the media coverage and many examples of workplace mobility, its population-wide extent is unclear (Felstead, 2012; Felstead and Henseke, 2017; Ojala and Pyöriä, 2018; Vilhelmson and Thulin, 2016): research in sociology and in organisation studies does not provide information about whether these new practices add up to a new urban economic geography. Only a few studies have explored whether workplace mobility is widespread. Ojala and Pyöriä (2018) perform a cross-sectional study of 30 European countries: they document that between 80% and 90% of workers still perform work at their employer’s premises at least several times a month, whilst also revealing that 40% to 60% also perform work from mobile locations, about 20% in vehicles and 10% in public spaces. Their panel data, covering a number of countries, does not inform on how the location of work has evolved over time, but the authors show that the degree to which employees are detached from their official POW is overstated.
Vilhelmson and Thulin (2016), in a study of telework in Sweden from 2005 to 2012, conclude similarly that:
most ordinary telework is performed in employee homes, while other places often highlighted in the literature (e.g. telecentres, telecottages, cafés and other public spaces) are much less frequent telework sites. This observation stands in contrast to discussions, suggesting that mobile locales could be an influential driving force of telework. [… However] we infer a potential for more teleworking in the near future. (2016: 93)
Felstead (2012) performs a longitudinal meta-analysis of 21 surveys conducted since 1981 in the UK. He concludes that ‘while work is being detached from conventional places of work, it is happening at a much slower rate than some claims suggest’ (2012: 31). Yet he also states that, despite patchy evidence:
it is clear that the future of work will be spatially diverse. Greater connectivity will mean that workers will be able to maintain a virtual presence wherever they may happen to be – on the train, in the car, a motorway service station or at home in the garden. (2012: 36)
More recently Felstead and Henseke (2017) have confirmed the slow but steady rise in remote work in the UK, whilst showing that it is not merely attributable to compositional factors (such as workers shifting towards more mobile occupations) but to more general changes in work location. These researchers also comment upon the limits of the data available to study the phenomenon: ‘the measure of remote working used is often based on surveys steeped in a tradition that sees a clear divide between home and work, with little in between’ (2017: 207).
Thus, although workplace mobility still appears to be limited, researchers expect it to grow: but they have identified limitations in the way work location is currently conceptualised and measured. It is therefore useful to extend the idea of ‘place of work’ so that urban geographers can incorporate workplace mobility into their analyses and empirically assess its possible impact on urban space. In the next section I propose an extension to the traditional idea of ‘place of work’.
A framework for studying the location of work tasks: Work location as probability space
In order to incorporate workplace mobility into conceptualisations and observations of the urban economy, it is useful to consider work as a set of related tasks (Brown and O’Hara, 2003): each task – such as sending emails, reading, programming, receiving goods, face-to-face meetings, using a machine – has its own location requirements, some of which resemble the requirements of other tasks, some of which do not. Furthermore, micro-work – quick checking of emails, rapid phone calls, brief diary verification – should also be considered when breaking down work tasks for the purpose of geographic analysis, since it can take place virtually anywhere.
During their working day (or whatever period is analysed), workers perform these tasks in different locations, spending different amounts of time on each of them. Instead of making the simplifying assumption that all work tasks occur at a single POW, work location can be approached as a probability space: for each worker/place combination there exists a probability that work will be performed. A worker’s probability space can be obtained by aggregating, over a certain period, the time spent performing tasks in each workplace. For the urban economy as a whole, the probability space of economic activity will be the sum of all workers’ probability spaces.
To make this concept tractable, a finite set of workplace types needs to be defined. This approach is used in probability theory, in which a set of all sample points (for example, the six possible outcomes of a dice throw) is identified and probabilities associated with each point are calculated (Itô, 1978).
In the context of this exercise, the set of workplaces from which the probability space is generated should reflect the types of workplace mentioned in the literature (Felstead, 2012; Ojala and Pyöriä, 2018; Vilhelmson and Thulin, 2016), starting with usual POW (W) and home (H) (Figure 1). To these should be added a series of other workplaces (such as co-working space, clients' premises, cafés … P1 to Pn) where the probability of performing work tasks is significantly different from zero. Tasks can also occur in transport and whilst on the move (T). And finally there is a residual (R) of other places where the probability of working is small.
Work location as probability space. (a) Classic conceptualisation of work location for a worker i.
For a specific worker i, each type of workplace they work from has geographic coordinates xy (xy could also be a neighbourhood or census tract), except for transport and for undefined residuals. For the city has a whole, three types of economic geography can therefore be analysed.
First, the probability of work tasks occurring can be summed across all workers by type of workplace– this provides information on the probability of work being performed at home, in a usual place of work, in co-working spaces, etc. This type of summation is analysed in the empirical section.
Second, the probability of work tasks occurring can be summed across all workers by geographic coordinates (or, more practically, by neighbourhood) – this would provide information about which neighbourhoods work activities are performed in, comparable with research that maps the location of employment across intra-metropolitan space.
Third, the probability of work tasks occurring can be summed across all workers, with subtotals by type of workplace and by coordinates (or neighbourhood): this would inform us of neighbourhoods where each type of workplace is used. It could uncover, for example, whether work in cafés occurs in particular residential neighbourhoods or whether work in co-working spaces is concentrated close to the CBD. This would be comparable with separately mapping workers who work from home and those who work from an establishment.
Figure 1 illustrates, for a single worker, the classic assumption (that economic activity is attached to a usual POW or to home), and a probability space such as outlined above. These are not antagonistic: the classic assumption is a special case of the proposed probability space
The rest of this paper has two goals. First, we frame our empirical analysis using census data to assess the degree to which POW has changed between 1996 and 2016: in doing so we document certain changes, but also reveal the limitations of standard data for such analysis. Second, we show how these limitations can be overcome by applying the idea of work probability space to exploratory data drawn from a survey designed to operationalise the concept.
Question, data and methods
To frame our exploratory study we explore whether it is reasonable to suggest that there has been a shift in type of workplace. The census POW question is analysed for data covering Canada’s ten largest urban agglomerations (CMAs), from 1996 to 2016. Given that most of the literature, especially case-studies and media reports, focuses on large urban areas, large cities have been chosen to ensure that small-town and rural behaviours do not bias the results. These data were ordered from Statistics Canada, with standardised geographies. 2
The census workplace question is formulated as follows: ‘At what address did this person usually work most of the time?’ 3 Four options are given: ‘Worked at home (including a farm); Worked outside Canada; No fixed workplace address; Worked at the address specified below’. This type of data has been used in most population-level studies of workplace and is fundamental to most empirical research on the geography of economic activity. It has, however, two key limitations for examining the changing geographies of work tasks. First, the category closest to mobile workplace is ‘No fixed place of work’– yet this category does not distinguish between people who perform all work tasks at a single location (whilst changing workplace on a weekly or monthly basis) and those who perform work tasks at different locations throughout the day. Second, the ‘worked at the following address’ category records where people usually work most of the time: it records an administrative attachment (combined with regular physical presence of unspecified duration and frequency) rather than the place where work tasks are actually performed: declaring a fixed work address does not preclude workplace mobility.
In the context of this paper the data therefore serve two purposes. First, and acknowledging their considerable limitations, they are used to present a population-wide impression of workplace change over a 20-year period and to detect possible shifts that may corroborate the hypothesis of increased workplace mobility. Second, and more fundamentally, their analysis starkly illustrates the need for a practical approach – one that is conceptually robust and that can reasonably be operationalised in surveys and the census – to gathering data on the geography of work tasks. The second part of the empirical analysis addresses this.
Indeed, the exploratory survey operationalises the concept of work probability space and shows how it can be applied. The survey is not representative: it is web-based, anonymous and only permits one response from any given IP address. As well as asking respondents descriptive questions (age, gender, work activity, city where they mainly work), it interrogates them on whether they perform work tasks in seven different types of workplace, reflecting those highlighted in the literature. These seven types of workplace constitute the work probability space’s sample points. They are:
`usual place of work outside the home (i.e. a factory, office, store or other place dedicated solely to work and where you work most of the time)’;
`at home’;
`in cafés, lobbies, restaurants …’;
`in transport, waiting rooms, airports …’;
`in co-working spaces, fab-labs, internet-cafés …’;
`at clients, suppliers, related establishments, collaborators …’;
`in other types of place’.
Respondents are asked to estimate the percentage of work time spent performing work tasks in each of these locations over the previous month (excluding vacations). Respondents can also provide comments, describe other places they work and mention whether they work during vacations and during time off. The survey concludes with questions about micro-work: whilst respondents are invited to consider this as they respond to workplace questions, it was felt that specific questions are also necessary since very short periods of micro-work may not appear in month-long estimations of time spent working from various workplaces.
The survey was active from mid-October 2018 to early April 2019 and was advertised on the website of the Centre de Recherches Interdisciplinaires en Études Montréalaises and in the newsletter of the Union des Municipalités du Québec. It was also circulated to acquaintances and colleagues. Thus, it is not a representative survey: rather, it partly utilises a snowball (or word-of-mouth) technique common in qualitative studies and partly dissemination through specific newsletters and websites. The respondents’ profile reflects these entry points. All but two respondents’ main region of work is a town (of over 50,000 people) or city but no geographic restrictions were introduced: placing a geographic filter upon responses was not considered useful since this survey is intended to illustrate a concept, not to represent a specific population. Given the non-random sample, none of the results can be extended beyond the survey’s 244 respondents, and the Chi2 statistic is used in a descriptive fashion to highlight the more relevant differences across a typology of respondents (Backhouse, 1984).
This type of non-probability sample ‘can be useful when the researcher has limited resources, time and workforce. It can also be used when the research does not aim to generate results that will be used to create generalisations pertaining to the entire population’ (Etikan et al., 2016). In particular, non-probability sampling can ‘confirm or displace anecdotal impressions’ (Skowronek and Duerr, 2009). In this case, the survey has two goals. Its first and principal goal is to illustrate how the concept of work probability space can be operationalised. Its second, subsidiary, goal is to provide insight into whether workplace mobility is widespread amongst respondents.
Where are work tasks undertaken?
The results section is divided into two subsections. The first frames the subsequent survey analysis by presenting some general trends derived from census data covering Canada’s ten largest cities: it establishes whether there is prima facie evidence that increasing numbers of workers work from non-traditional workplaces and illustrates the limits of this type of data. The second section is a ‘proof of concept’, showing how work probability spaces derived from survey responses can provide better insight into workplace mobility.
What do census data reveal about workplace mobility?
Table 1 reports responses to the Canadian census question ‘At what address did [the worker] usually work most of the time?’ in the week preceding the census. Between 1996 and 2016, the proportion of workers declaring work at a fixed address outside the home (that is, at a usual POW) decreased by 5.4%, whilst the proportion declaring no fixed POW increased by 4.2%. The proportion working at home has remained stable, increasing by only 1% over the period.
Where does the worker usually work most of the time? Census data.
Source: Statistics Canada, special tabulations of census data, 1996, 2006, 2016. Some small classes (such as family workers), as well as most sectors, have been excluded, so totals do not all add up.
The trend is not reproduced identically in all sectors. For KIBS (Knowledge Intensive Business Services), whilst the proportion of workers with a usual POW has decreased, this translates into more working from home and more working outside of Canada. In Educational Services there has been a marked increase in workers with no fixed POW, accompanied by a modest rise of people working from home. Workers in Public Administration overwhelmingly have a fixed POW, whilst a small but increasing proportion of workers has no fixed POW. Finally, when Other Services are considered – a sector dominated by cultural and information industries but also including utilities – the proportion of workers with no fixed POW rises from 18.6% to 21.6% between 2006 and 2016.
In this selection of sectors – which are chosen because they correspond to the sectors in which survey respondents work – fixed POW is decreasing (except for Other Services, where it is stable). Of note is the fact that, for KIBS, it is work at home rather than no fixed POW that has increased. These results corroborate those of Ojala and Pyöriä (2018), Felstead (2012) and Vilhelmson and Thulin (2016) in that a considerable majority of workers in Canadian cities still report a usual POW. However, these data do not actually pick up the variety of locations from which work-related tasks are performed during the day, one of the key elements of workplace mobility: although our data reveal a population-level shift towards workers declaring no fixed POW, with certain sectors displaying higher levels of such responses than others, 4 the shift does not map easily onto the concept of workplace mobility.
Furthermore, the population-wide shift towards no fixed POW is marked (in absolute terms the number has more than doubled) but not overwhelming (the vast majority of workers still declare a usual POW). It is also worth noting that the rate of increase has slowed, from 2.6% between 1996 and 2006, to 1.6% in the following decade, and that no fixed POW actually decreases for KIBS and Other Services.
These census data can be analysed in a variety of ways: for example, workplace shifts by occupation and employment status can be explored, as shown in Table 2. Across these dimensions no fixed POW systematically increases between 1996 and 2016, and it is amongst manual professions, trades and the self-employed that no fixed POW is most prevalent – again corroborating Ojala and Pyöriä (2018) who show that it is currently ‘traditional’ workers who are most mobile.
Descriptive statistics of survey respondents.
Notes: n = 246.
One person declared being neither male nor female. This person has been classified as female for the purposes of analysis.
‘Other’ principal towns or cities of work were in the UK, the rest of Europe, Australia or the USA.
Even though our results corroborate evidence drawn from studies that use more appropriate (though not easily available) data, they elide a central issue: they rest upon census data that assume work tasks either take place in a single location (whether home or workplace) or in no fixed location. Census data therefore hint at, but do not measure, workplace mobility. This has become an issue in light of mounting case study, survey and anecdotal evidence that work tasks occur in a wide variety of locations. The next section presents the results of an exploratory study which operationalises the concept of work probability space.
Work probability spaces: An illustration
The online workplace survey gathered 246 valid responses. Respondents are from four main sectors: Education, KIBS, Public Administration and Other (with a predominance of cultural sector workers but also a few people working in utilities, health and manufacturing). Table 3 summarises their responses to the POW survey question, worded similarly to the census question. 5 Comparing these responses with the census (Table 1) provides a link between the non-representative survey and the census.
Where does the worker usually work most of the time? Survey data, by sector.
Whilst the survey responses differ in level from the census, they are consistent with it: with respect to usual POW, only education diverges substantially and that is probably because the survey mainly picks up university workers (by virtue of how the survey was disseminated). It is probable that a higher proportion of professors work from home than college or high-school teachers, and sessional lectures and post-doc researchers may declare no fixed POW. Fewer surveyed KIBS workers work from home than do so in the census, with more declaring no fixed POW. Generally, proportions of survey respondents declaring no fixed POW are commensurate with the census: whilst levels differ, the ranking of sectors in both the survey and the census are the same.
Were these data to be mapped in order to analyse the urban space economy, 100% of the work tasks performed by people declaring a ‘usual POW’ and ‘at home’ would be assigned to these workplaces and the tasks of those declaring no fixed POW would be unassigned geographically. Thus, about 84% of the respondents would be allocated a POW, and our understanding of the space economy would proceed on that basis, as it does when census-type data are used.
If work location is conceptualised as a probability space, then it produces different mappings. Rather than assigning each worker either 100% to a specific workplace (home or usual POW) or 0% (because they have no fixed POW), each worker’s tasks are assigned to various workplaces in proportion to the time spent working there. The exploratory survey implements this alternative approach to establishing workplace location. After describing to respondents what is meant by ‘work task’, 6 they are asked to evaluate what percentage of their work time, over the last month, was spent working in each of seven different types of workplace. In effect, for each worker a seven-point probability space is defined and each has his or her own probability profile corresponding to the proportion of work time spent at each of the seven points.
To aggregate these data, hierarchical cluster analysis is performed on the time (or probability) profiles in order to construct a typology. A ten-cluster solution is found, 7 with two distinct branches. One branch gathers clusters 11, 12, 20 and 23 (see Table 4): these represent 73.4% of respondents, who have in common that 80% or more of their time is spent working at a usual POW (50% or more) or at home (10–30%). The other branch comprises five small groups of workers, who have in common less than 38% of time spent at a usual POW, with the rest spent in a variety of other work environments: these five clusters are grouped into the ‘JOIN’ cluster since it is not possible to meaningfully analyse clusters with small membership.
Profiles of workers by % of time spent working in seven different types of locale.
Note: UPOW: usual place of work.
Five profiles emerge:
CL20: 81 respondents (32%) predominantly work at their usual POW, working 8% of the time from home and 2% from a variety of other workplaces.
CL11: 49 respondents (20%) work 73% from their usual POW, 13.5% from home, and also in transport (6.9%) and in cafés (2.4%).
CL23: 17 respondents (7.0%) have a profile similar to CL11 except that they do not work in transport locations but work in a variety of workplaces not covered in the survey.
CL12: 32 respondents (13%) spend just over 52% of their work time in a usual POW, 33% of their work time at home and 8.9% in cafés.
JOIN: this heterogeneous group of 65 respondents (26.6%) comprises: (i) a group of 37 (which dominates the JOIN cluster) who work from home 74% of their time; (ii) 6 who work 47% from home, 19% in transport, 8% in co-working spaces and 14% in a usual POW; (iii) 8 who work 65% of the time from co-working spaces; (iv) 7 who work about 30% of the time from each of a usual POW, home and clients' premises; (v) 5 who work over 50% of the time from other workplaces; and (vi) 2 who work 85% of the time in cafés.
A feature of these clusters is that home and usual POW are key anchor points. Whilst work tasks are undertaken in a variety of workplaces, all but about 20 respondents (in the JOIN cluster) work predominantly from a usual POW or from home. If the time of the 244 respondents is weighted equally and aggregated to estimate the work probability space for all survey respondents, there is a 60.3% probability of work occurring at a usual POW, 24.6% at home and 15% from a variety of other workplaces (Table 4, row titled ‘Mean’).
The aggregate profile – which would reflect the urban economy were the sample representative – can be mapped onto the work probability space in Figure 1 using the following numbers:
W% = usual place of work = 60.3%
H% = at home = 24.6%
P1%=cafés, lobbies, restaurants etc.=3.8%
P2% = co-working spaces = 3.1%
P3%=client's or in related establishment= 1.4%
T% = in transport = 2.8%
R% = elsewhere = 4.0%
In this study only types of workplace can be analysed: the xy coordinates (or neighbourhoods) for each workplace were not obtained from respondents.
What distinguishes workers who report different work probability spaces?
The five different probability space profiles do not differ in terms of the age, occupation, education level, city (or region) of work or gender of their members 8 (Table 5). The similarity of work probability spaces across age is particularly relevant: amongst the respondents there is no difference between digital natives, millennials and older workers. There are differences, though, across other dimensions (Table 5).
Clusters (that is, work probability spaces) by worker characteristics.
Notes: CL11: UPOW, transp. & various; CL12: Home, UPOW, cafés, various; CL20: UPOW; CL23: UPOW, Other & various; JOIN: All workplaces, esp. home.
The Chi2 test is used as a descriptive tool in order to highlight which dimensions are more strongly connected to cluster membership than others amongst the 244 respondents. Results cannot be extended beyond these respondents (Backhouse, 1984).
Whilst more self-employed respondents work from home or from other workplaces (CL12 and JOIN), being an employee does not preclude the possibility of working from a variety of different places: 48% of the JOIN members and 75% of CL12 members are employees. The JOIN cluster also stands out in terms of establishment size – fewer members work in large organisations, with more declaring that they do not work in any establishment – that is, that they are freelance or independent. Most workers have a fixed workstation, though fewer in the JOIN cluster do: hot-desking and desk-sharing is a limited phenomenon amongst respondents.
Where work tasks are performed is connected to economic sector: the most ‘classic’ workplace profile, CL20 – with 90% of time spent at a usual POW – gathers 46% of KIBS and of Public Administration respondents. 9 Whilst these sectors display a more traditional workplace profile, 8% of Public Admin and 29% of KIBS workers perform activities from a variety of places (JOIN). The JOIN cluster is, however, dominated by workers classified as ‘Other’, mainly in arts, hospitality and health and social services.
Respondents with workplace profiles CL11, CL12 and CL20 tend to work closer to downtown (their usual or latest POW is within 5 km of the CBD), workers in CL23 (who spend considerable time at clients' premises) and JOIN (who work from a wide diversity of places) tend to be more suburban. This is worthy of note, especially with respect to the JOIN category: whilst co-working spaces and mobility are often associated with downtowns, the opposite seems to be the case in this survey – people operating from a greater diversity of workspaces tend to work more in suburban locations. This may reflect that, notwithstanding this locational diversity, much time is spent working from home. Since home is the base for many people in the JOIN category, their alternative workplaces may complement their home base: in the open questions (which will not be analysed in detail in this paper) a number of people who principally work from home mention the isolation and boredom this elicits and speak of the need ‘for a certain place where [they] can interact and socialise with people in a similar situation’ (ID: 379), since working at home ‘can be depressing’ (ID: 331).
Respondents in clusters where people work more from home (CL12 and JOIN) tend to live closer to their usual (or latest) POW, with over 25% declaring this distance to be less than 1 km. Members of CL12 are the least likely to travel over 7 km to work: this, combined with the central location of their usual (or latest) POW, suggests that the CL12 workplace profile gathers people who live and work in downtown areas. In all other clusters – including JOIN – at least 35% of respondents travel more than 7 km to their usual (or latest) POW.
Respondents were asked about micro-work – defined in the questionnaire as ‘emails, phone calls, texts, diary …’– which can occur in any type of place, on the move, often during interstitial time snippets. About 10–15% of respondents perform micro-work whilst with family and friends, and almost one-quarter of respondents with the JOIN workplace profile do so. Micro-work is more common during the working day, when respondents are ‘between places, eating or waiting’: about 40% perform micro-work in these conditions, except those in CL20, for whom micro-work is less common.
Finally, certain respondents give examples of other places they perform work: some mention working in national parks (ID: 273), whilst walking the dog (ID: 399), as a patient in hospital (ID: 407) or in a hammock in the park (ID: 445). Amongst respondents such workplaces are the exception rather than the norm.
Discussion and conclusion
The increasing ease with which work tasks can be performed from a wide variety of workplaces (workplace mobility) has been much discussed in the media and amongst scholars since the late 2000s. In this paper the limits of standard workplace data are discussed and illustrated and a framework is proposed (and applied) that allows workplace mobility to be studied using tools and concepts that resemble, but extend, those currently used to study the urban space economy. In particular it is suggested that the heuristic short-cut used in many analyses and databases – assigning each worker to a particular POW – is a particular instance of a more general approach to conceptualising work location. This general approach consists in considering work location as a probability space, the sample points of which consist in workplaces (or locations) selected because of their theoretical or empirical relevance.
Using responses to an exploratory survey, we show how probability space can be operationalised to explore workplace mobility. Given the simplicity of the questions asked about a finite number of workplace types (similar to those of the ECWS survey used by Ojala and Pyöriä, 2018), this approach can be deployed in larger-scale surveys (or in the census) provided that agreement can be reached regarding major types of workplace (i.e. sample points). It would be more difficult, though not impossible, to enquire about the geographic coordinates of each type of workplace by, for example, asking for the address of the most recent one used.
However, a key question is the extent to which workplace mobility has become widespread. It has been much discussed both in the business, technology and sociology journals, as well as amongst researchers interested in mobility, but few studies document how it has changed over time. Business studies focus on workplace organisation, on real-estate and on the way flexible or agile work practices attract millennials and enable cooperation and innovation. Technology and sociology research focuses on the ways ICT enables new workplace practices. Mobility scholars study mobility under its numerous guises.
Although this research describes interesting and new phenomena, its results are not intended to (and cannot) be generalised to the scale of the city. From the perspective of the urban spatial economy, workplace mobility is relevant if it meaningfully impacts how the space economy functions. Whilst acknowledging the data’s limitations, our census data show that between 1996 and 2016 there has been a small decrease in the proportion of workers declaring a usual POW (from 85.5% to 80.1%), and a relatively large (but absolutely modest) increase in the proportion of those with no usual POW (from 7.8% to 12.0% – with slightly slower growth between 2006 and 2016). The proportion of workers declaring home as their principal workplace has remained virtually unchanged. This evidence rests upon data that do not capture workplace mobility but are consistent with the idea of modest – but not (so far at least) revolutionary – change. 10 It is consistent with Felstead (2012), Felstead and Henseke (2017) and Vilhelmson and Thulin (2016) who document a slow but consistent change in where people work.
Whilst it is useful to integrate workplace mobility into concepts and methods used to study work location, evidence gathered and reviewed in this paper suggests it is not yet a widespread phenomenon. Studies that analyse the location of economic activity by assigning each worker to a unique POW remain good approximations of where economic activity actually takes place across metropolitan space. In our exploratory survey, 85% of survey respondents’ work time is spent either at home or at a usual POW – so assigning 100% of work time to home or usual POW is a reasonable approximation, all the more so because aggregating the (fixed) locations of many workers will smooth stochastic imprecision. However, conceptualising work location as a probability space adds to our understanding of cities and will enable us to better capture changes: indeed, another way of considering the same data is that, amongst respondents, 15% of work time (just under one day a week for a five day week) is spent in non-traditional workplaces. This may not revolutionise our basic understanding of the connection between the economy and urban space, but is a phenomenon that cannot be ignored (Felstead and Henseke, 2017).
Like many changes, especially those that showcase technological innovation and ICT, workplace mobility has been hyped; yet the diminishing rate of increase of ‘no usual POW’ hints that the last two decades are not necessarily indicative of future development. Fast change occurred as internet and cell-phones first became widespread in the 1990s and early 2000s: the period following the introduction of smart phones (introduced in 2007) witnessed slower change in Canadian cities. As some of the challenges of workplace mobility – such as coordination, team work, quality of communication, concentration, privacy, confidentiality, loneliness, over-stimulation – become evident (Felstead and Henseke, 2017; Hekkala et al., 2017; Keeling et al., 2015; Pochepan, 2018) its growth may further slow. Workplace mobility will not disappear and probably will not decrease: rather, it may plateau as organisations and workers take stock of new workplace possibilities and determine whether they are suitable. For the time being, working from home or from a usual POW seem to be adequate for most workers most of the time: but, in order to state this with confidence and to understand change, better data are required. This paper proposes a straightforward conceptual framework, and an example of its operationalisation, that can serve to gather and analyse such data.
Footnotes
Acknowledgements
I would like to thank Mario Polèse, Filipa Pajevic and Lukas Stevens for reading and discussing earlier versions of the paper, as well as the Union des Municipalités du Québec and the Centre de Recherches Interdisciplinaires en Études Montréalaises for helping me circulate the survey. Thanks also to the three reviewers who helped improve the paper. I remain solely responsible for its content.
Funding
The author received no financial support for the research, authorship, and/or publication of this article.
