How to attract employees back to the office? A stated choice study on hybrid working preferences

It is expected that most office workers will work from home more often after COVID-19, but it remains unclear who is inclined to go back to the office and who is not when hybrid working becomes reality. Existing studies lack insights how (design) characteristics of the available office and home workspace influences workspace choice behavior. This survey-based stated choice experiment identifies two employee segments: one that intends to re-embrace the office and one that prefers to stay home a lot. Especially the expected crowdedness on the floor and availability of private spaces for concentration and meetings determined these employees ’ choices. Also, the office workers segment contained relatively more male, highly educated, full-time workers with communication as an important component of their work and a short commute, while the home workers segment contained relatively more females, part-time employees, and administrative roles, plus employees with more individually focused work and a long commute.


Introduction
The past year, the COVID-19 pandemic allowed office organisations and employees to get acquainted with working from home more than before. For many workers, this has been a positive experience regarding perceived improvement of work efficiency, control, and work-life balance (Ipsen, Van Veldhoven, Kirchner, & Hansen, 2021). Recent studies (e.g. Marzban, Durakovic, Candido, & Mackey, 2021;Naor, Pinto, Hakakian, & Jacobs, 2021) show a larger inclination to work from home more often after the pandemic, and many organisations are voicing the aim to let people work in a hybrid way. Halford (2005) was one of the first to define the term hybrid work as those who work both at home and at their employer's sites. But not until the Covid-19 pandemic it became a highly popular term. Until recently, the term teleworking covered both employees working fully from home and/or those working in a hybrid way in scientific studies. Hybrid work was only referred to in the context of teams working with co-located and virtual members (e.g. Deshpande, Sharp, Barroca, & Gregory, 2016), or people with marginal part-time jobs or self-employment (Mailand & Larsen, 2018). As Halford (2005, p. 20) already mentioned: "Being employed to work both at home and also in an organisational setting, using ICTs to maintain workloads and relationships across both domestic and organisational spaces raises new questions that lead beyond the sum of existing debates about teleworking and virtual organisation." Despite an interest in hybrid work, it remains unclear which employees want to remain working from home and who prefers to go back to the office as much as possible, and to what extent office design and the available home workspace influence this choice. Older studies on the inclination to work from home partially (e.g. Baruch & Yuen, 2000) showed a wide distribution in willingness to work from home and have called for more research. The gained experiences of many employees during Covid-19 lockdowns might have changed their opinions. So far, research on where people prefer to do their work remains scarce and has not addressed yet how office and home workspace design will influence preferences and thus work location choice behavior after the pandemic. Hybrid working inclination studies focussing on the situation after Covid19, only ask whether the home working experience was pleasant but ignore the quality of the available office workspace as an alternative (e.g. Barrero, Bloom, & Davis, 2021, pp. 2020-2174Nguyen, 2021). The choice between office and home will not only depend on workspace preferences, but also on the type of tasks planned for the day (concentration, communication, or a mixture of both). As argued for the activity-based office (Eismann et al., 2021), hybrid working means that employees must plan their workday in advance and carefully select the most appropriate workspace (office or at home) based on activities. Given the reluctance to switch between types of workspaces inside activity-based offices (Hoendervanger, De Been, Van Yperen, Mobach, & Albers, 2016), it remains unclear whether employees will let their planned activities determine where they decide to work. So far, office design studies about supporting both concentration and communication activities are scarce and do not include the option to work from home. They are generally focused on measuring either perceived concentration or communication at the office, or they use concentration and communication as indicators of workplace satisfaction or productivity (Brunia, De Been, & van der Voordt, 2016;De Been & Beijer, 2014;Van der Voordt, 2004) or vitality (Wohlers, Hartner-Tiefenthaler, & Hertel, 2019). Studies on hybrid working, on the other hand, largely ignore office design quality and focus only on home workspace suitability and on commuting to work (e.g. Nakrošienė, Bučiūnienė, & Goštautaitė, 2019;Ollo-López, Goñi-Legaz, & Erro-Garcés, 2020).
The aim of this study is therefore to identify which office and home workspace characteristics determine where people decide to do their communicative and/or concentrative work after the pandemic: in the office or at home. In addition, it identifies employee segments with specific workspace location preferences and explains them by personal, work-related, and home workspace characteristics. Insights show how offices should be designed to attract people that are inclined to work from home. In addition, they will show who can be expected at the office and who not when given a choice.
Concentration and communication are important basic needs for office workers (Deci & Ryan, 2014;Lieberman, 2013;Ryan & Deci, 2001;Vischer, 2008). However, they are two extreme contrary requirements that demand a well-functioning and balanced office environment (Roper & Juneia, 2008). Concentration work requires the ability to select information for a specific goal, and at the same time ignoring distractions from outside (Schwartz & Kaplan, 2000). Being distracted in the work environment can have a negative impact on employees' workplace satisfaction and work performance Roper & Juneia, 2008;Sundstrom et al., 1994), concentration (Banbury & Berry, 2005) and eventually health (Colenberg, Jylhä, & Arkesteijn, 2020;Seddigh, Berntson, Bodin Danielsson, & Westerlund, 2014). Performance deteriorates especially from arousal when involved in more complex tasks, as shown by the Yerkes-Dodson law (Yerkes & Dodson, 1908). Also, far more office work activities require higher concentration levels than employees expect (Hoendervanger, Le Noble, Mobach, & Van Yperen, 2015). Working from home is especially beneficial when working individually (Bentley et al., 2013;BodinDanielsson & Theorell, 2019;Hamersma, De Haas, & Faber, 2020;Van Dyne, Kossek, & Lobel, 2007). So, employees with mostly individual concentrated work might prefer to avoid the office with its noise distractions when hybrid working becomes common.
Nonetheless, humans are fascinated by other human beings, so communicating with each other is biologically programmed (Csikszentmihalyi, 1990). Communication is expressed in multiple terms (collaboration, interaction, knowledge-sharing) and aims to transfer information from one person to another that has meaning to the participants (Culnan & Bair, 1983). Communication at work is both planned and unplanned. Pre-Covid studies show that working from home may lead to negative experiences due to a decreased number of unplanned communication (Derks, Agterberg, Beumer, & Weel, 2011;Peters, Van den Dulk, & Van der Lippe, 2009;Vos & Van der Voordt, 2001). Studies during the pandemic showed that all collaboration became technology based and therefore more orchestrated and planned, which again had negative effects on communication (Waizenegger, McKenna, Cai, & Bendz, 2020). So, planned or desired unplanned communication activities might drive people to go to the office anyway despite hybrid working policies.

Workspace characteristics to support communication and concentration
Not only cognitive reasoning, but also the physical office design plays a role in communication and concentration behavior (Wohlers & Hertel, 2016). Existing research has provided evidence for several design aspects that influence experienced support of both communication and concentration. First, many studies have shown that people still are dissatisfied with office noise (e.g. Brennan, Chugh, & Kline, 2002;De Croon, Sluiter, Kuijer, & Frings-Dresen, 2005;Hedge, 1982;. In offices with high noise levels, employees' memory performance and motivation decrease while tiredness increases (Jahncke, Hygge, Halin, Green, & Dimberg, 2011). In case of speech/conversations, it is not so much the level or duration of the sound but the intelligibility that disturbs others (Banbury & Berry, 2005;Pierrette, Parizet, Chevret, & Chatillon, 2015;Sundstrom et al., 1994). Hongisto (2005) found that especially performance on complex tasks (concentrated work) decreases with 7% when unattended speech is highly intelligibility, while there were no effects when speech intelligibility was low. On the other hand, when the conversation is intelligible, the awareness of the other person increases the number of spontaneous interactions (e.g. Appel-Meulenbroek, De Vries, & Weggeman, 2017).
Second, Social Interference Theory (SIT; see Davis, Leach, & Clegg, 2020) points out that configurational features of office design affect unexpected encounters. A more open workspace may increase communication due to increased visibility, but also increases the chance of being distracted by others (De Croon et al., 2005;Rashid, Kampschroer, Wineman, & Zimring, 2006). In addition, regarding crowdedness two variables are relevant (Dean, Pugh, & Gunderson, 1975): "space size" and "number of people". Physical proximity in open spaces (how close employees' desks are located to one another) determines levels of communication and distractions when trying to concentrate (Davis et al., 2020). In addition, having sufficient space can be seen as a fundamental human need (Stamps, 2009). The number of people that are present (occupancy rate) is associated with concentration difficulties (Pejtersen, Allermann, Kristensen, & Poulsen, 2006) but when you can see more people behind your desk, unplanned interactions increase (Appel-Meulenbroek et al., 2017).
Another aspect is the position of the desk in relation to walking routes. Most of the movement in buildings is concentrated along corridors and entrances of areas, which bring people past other people's desks (Penn, Desylas, & Vaughan, 1999). People located near such high-traffic areas are more likely to interact with people passing by (Davis, 1984). However, the layout of the office and how routes and walls are positioned also influences the extent to which people experience visual distractions (Gielissen et al., 2018). In general, people find it distracting when others are moving and walking around them (Bell, Greene, Fisher, & Baum, 2001).
Another two relevant office design aspects are the availability of additional spaces for either concentrative or communication activities. It is important to facilitate the office with enforced areas of quietness where employees can move when they require a high level of concentration (Banbury & Berry, 2005). Especially employees without assigned desks are dissatisfied with the office workspace when access to such rooms is poor (Bodin-Danielsson & Theorell, 2018). These rooms create more privacy and provide more control over the environment to regulate interactions (Kupritz, 1998). They also provide possibilities to have confidential conversations, which is important for the communication flow in the office (Steelcase, 2002). Although such enclosed rooms give a sense of privacy, they decrease the chance that employees see each other and start conversations.  called this the "privacy-communication" trade-off.

Personal characteristics and choice behaviors
Personal characteristics (demographical, work-related, and home workspace-related) influence workplace preferences and choice behaviors too. For example, Awada Lucas, Becerik-Gerber, and Roll (2021) found that female and older workers felt more productive while working from home during the pandemic than their counterparts and are more inclined towards remote work. Gender also relates to communicative behavior, where men have fewer face-to-face interactions at their desk than women do (Weijs-Perrée, Appel-Meulenbroek, & Arentze, 2018). Regarding age, preferences differ in commuting and for example the adjustability of the indoor climate (Rothe, Lindholm, & Nenonen, 2012). In addition, the level of education is related to the importance people give to workplace design and communication (De Been & Beijer, 2014) and their propensity to work from home as well (Ollo-López et al., 2020). Personality is another variable to include for differences in preferences (Hartog, Weijs-Perrée, & Appel-Meulenbroek, 2018;Bozioneles, 2004;Oseland, 2009). Last, young children at home can cause distraction for the employee while working from home, and this will affect their ability to work (Soetman, 2011).
Regarding work-related characteristics, long-tenured employees find adaptation more difficult than employees who work at the company for a shorter period (Fossum, Arvey, Paradise, & Robbins, 1986). They might thus have different preferences if hybrid working is new to them. In addition, the position in the organisation could determine preferences, as it relates to time spent on certain activities while in the office (Brill & Weideman, 2001). In addition, the amount of work hours per week (De Been, Van der Voordt, & Haynes, 2016) and how people spend this time at different types of workspaces when they are in the office (Greene & Myerson, 2011) influences workspace preferences. Last, the probability of hybrid working increases with commuting distance (Helminen & Ristimäki, 2007;Ollo-López et al., 2020).
A third category of personal characteristics is the physical workspace that one has available at home. Office workers might not have a dedicated workspace or desk at home (Hill, Ferris, & Märtinson, 2003). If employees have a separate room at home, they can more likely avoid distractions (Ng, 2010). At home, noise from doorbells, visitors, or telephones ringing, conversations between household members, or sounds from televisions might be distracting (Jensen, 2007;Ng, 2010). Also, employees who share their home-workspace are more frequently distracted (Awada, Lucas, Becerik-Gerber, & Roll, 2021) and people usually do not have adjustable tables, ergonomic chairs, or external monitors at home either (Davis et al., 2020;Janneck, Jent, Weber, & Nissen, 2018). All these distractive aspects are likely to influence inclinations towards hybrid working. In addition, an indication of the amount of workspace someone has available at home for work might explain work location choices. During the pandemic, a Spanish study showed that 31% found their home workspace to have an inadequate size and considered it less suitable for work because of that (Cuerdo--Vilches, Navas-Martín, March, & Oteiza, 2021).

Stated choice experiment
This study uses a stated choice experiment to describe and predict workspace preferences and choice behavior given specific communication and/or concentration work activity scenarios for a workday. In stated choice experiments individuals are presented with hypothetical alternatives that are described by a number of characteristics (attributes) and asked to choose the alternative they like most given a specific context (Hensher, Rose, & Greene, 2015). Setting up a stated choice experiment involves a number of steps (e.g., Hensher, et al., 2015): selection of influential attributes; specification of relevant attribute levels; selection of experimental design; constructing the choice task; data collection procedure, and model estimation. In our study, employees are asked to choose their preferred work location given a set of work activities they need to perform during that day.
First, the attributes and the levels used to describe the hypothetical workspaces are selected based on the literature review (see Table 1). . Subsequently, the experimental design is selected to create the hypothetical alternatives. A full-factorial design consists of all possible combinations of attribute levels and allows independent estimation of all main effects and all two-way and higher-order interactions. However, combining all levels into workspace alternatives provides 3 7 = 2187 possible combinations. Hence, researchers often use fractional factorial designs that allow the estimation of at least all main effects while giving up the possibility of estimating some interactions and which leads to a significantly lower number of hypothetical alternatives. Therefore, in this study, an orthogonal fractional factorial design consisting of 18 alternatives, that allows independent estimation of all main effects, was used to create the hypothetical workspaces. Then, the hypothetical alternatives were randomly placed in choice sets, where each choice set consisted of three hypothetical workspaces and the (alternative) answer option "working from home".
Also, three different activity-scenarios were included to deal with Table 1 Attributes and their levels. Not directly next to a walking route different hybrid working preferences when being faced with different type of work activities (communication-scenario, concentrationscenario, 50/50-scenario): A. Imagine that you have a workday where you have many (un)planned meetings and relatively little individual (concentrated) desk work that day. B. Imagine that you have a workday where you have a similar amount of time spent on (un)planned meetings as on individual (concentrated) desk work that day. C. Imagine that you have a workday where you have a few (un)planned meetings and a relatively large amount of individual (concentrated) desk work that day.
In the survey, each respondent was confronted with three times three different work activity-scenarios and choice sets (3xA, 3xB, and 3xC = 9 choice sets in total. Furthermore, the survey explains that respondents need to imagine that the COVID-19 pandemic will no longer affect the work situation in the office and that working from home is allowed by free choice where to work each day (see Fig. 1 for an example).
The survey also asked the demographical (age, gender, education level, personality, household composition), work-related (tenure in the office, position in organisation, working hours, distribution working hours per location before COVID-19, distribution working hours per work activity before COVID-19, importance of work needs, travel time to office), and home workspace-related questions (home workspace setting, presence of other people, size of home workspace, noise at home, available furniture) that came forward from the literature review.

Modelling approach
First, a multinomial logit model is estimated which assumes that all the estimated parameters are equal for all employees; the utility for employee i for workspace j on workday t is calculated as follows: where X ijt represent all characteristics of the workspaces with their relative weights, parameters β ′ , to be estimated. The error term, εi jt , represents the unobserved heterogeneity.
However, as heterogeneity in workspace preferences can be expected, a latent class model is estimated. Simultaneously the employees are segmented and a separate set of utility parameters for each of the segments (S) is estimated for each of the work activity scenarios. Then the utility function for employee i's choice among j workspaces on day t, given that employee i belongs to a latent class s, (s = 1, …, S), is: s is a segment-specific parameter vector. The probabilities of choice can be derived from the utility function and for each segment, the probability that employee i chooses workspace j on workday t is: Segments are considered unobserved (latent) classes and have certain segment probabilities as follows: Z i is an optional set of observable individual choice situation invariant variables. In this study however, we do not include such variables in the model estimation directly but use a two-step approach. First each employee is assigned to the latent class with the highest probability and subsequently, the differences between classes based on a number of socio-demographic and other characteristics is measured using crosstabulations (with Chi-square tests) and analysis of variance (with F ratio). The latent class parameters are estimated using maximum likelihood estimation. Madden's Rho 2 provides an indication of the model fit and to compare how the number of classes changes the model performance the minimum Akaike Information Criterion or AIC (Akaike, 1987) is used. Fig. 1. Example of the stated choice experiment question with choice task.

Results
The survey was sent to 470 office workers of a commercial real estate company (in January 2021) of which 218 respondents fully completed it (response rate = 46%). A description of the sample and case organisation on a number of characteristics is shown in Table 2. More men (58.7%) than women (41.3%) participated, with average ages of 39 and mostly highly educated respondents (80.3%) that on average worked 37.5 h a week (contract hours). Regarding gender, age, and working hours the sample was representative for the whole organisation.

Workspace preferences per work activity scenario
Based on the workspace choices made by the respondents in the experimental choice situations it was estimated which characteristics were important for specific work activities (using Nlogit6). In the estimation, the constant was coded as 1 for the workspace alternatives, and 0 for working from home. All other attributes were effect coded (1, − 1). First a multinomial logit model was estimated including separate parameters for all work activity scenarios, see Table 3 for the estimated parameters. The model fit was reasonable with a McFadden Rho-squared of ρ 2 = 0.13.
As expected, constants, for the communication-scenario (β = 1.050) and for the 50/50-scenario (β = 0.297), indicated that respondents preferred to work at the office (see Fig. 2). The estimated constant for the concentration-scenario (β = − 1.344) however showed a negative value meaning that respondents who mainly need to perform concentration activities during the day preferred to work from home. Using the MNL model, the probability was calculated, that a respondent chooses a specific work location, office versus home, given a certain work activity scenario. This resulted in a 74% chance that employees chose a workspace at the office over working from home when they had a workday with mainly communicative work and 57% when the workday consisted of both concentrative and communicative activities. However, for a day with concentrative work, the chance is high that employees prefer to work from home (79%).
The MNL model output also included preferences (estimated βs) for the various workspace characteristics per activity-scenario, as shown in Fig. 3. Note that if an estimated parameter is not significant it is assumed to be 0, thus not affecting the workspace choice, and therefore not shown in the figure. For the concentration-scenario, the attribute levels of openness, space size, and availability of communication spaces did not influence the workspace choices significantly.
First, based on these parameters the relative importance of the workspace attribute preferences per activity-scenario was calculated (see Fig. 3). In all activity-scenarios, the position of the desk in relation to walking routes, crowdedness and noise were important in deciding whether to work at such an office workspace. On the contrary there were no specific strong preferences for the openness of the office in any activity-scenarios. For the communication-scenario, crowdedness and the availability of communication spaces were relatively most important. For the 50/50-scenario, this was the noise-level and the position of the desk in relation to walking routes. For the concentration-scenario, crowdedness, noise-level, and the availability of concentration spaces were most important.
Secondly, there are workspace conditions (quality levels of the attributes) that were not preferred in any of the activity-scenarios (see Fig. 4): • A workspace surrounded by regular intelligible conversations, • a busy floor where almost all surrounding desks are occupied or • an isolated workspace not directly next to a walking route.
In the concentration scenario employees preferred a workspace that is not too disturbed with ample availability of specific concentration spaces to withdraw too. However, a quiet, isolated (from routes) enclosed workspace was not preferred at all. For all three scenario's a lack of available concentration spaces was a clear no-go. Another interesting finding was that even on days with lots of communication, people still did not prefer to hear intelligible conversations at their desk or to work on a busy floor. They expected abundantly available dedicated spaces to have interactions at, so not necessarily near their desks.

Segments with different workspace preferences
The latent class model estimation resulted in two different latent segments. The Rho-square (ρ 2 = 0.19) indicates that the model fit was fine. Note that, as the MNL model shows quite similar preferences per activity-scenario for openness, space size, crowdedness, and the position of the desk in relation to walking routes, these attributes were assumed to be equally preferred by the respondents across all activity-scenarios, and therefore no activity-scenario effects were considered for these attributes in the LCM model. Segment 1 consisted of 69% of the respondents and segment 2 of 31%. The main difference between the two segments was the constant parameter. When looking at the constant values from segment 1, see Table 4, it can be concluded that this segment had an overall preference for a workspace in the office, except in the concentration-scenario. Therefore, segment 1 was labelled "Office-workers". For segment 2, the constant parameters showed different outcomes; specifically, all values were negative. This means that segment 2 favoured to work from home in all activity-scenarios. Therefore, segment 2 was defined as "Home-workers".
When looking more specifically at the parameter estimates, in general they were higher for the Office-workers than for the Home-workers. This is because the Home-workers segment had an overall preference for working from home. Furthermore, both segments preferred a workspace that has some unintelligible conversations/sounds. A quiet workspace (noise) was only preferred by the Office-workers in the concentrationscenario. In general, the Office-workers preferred a partially enclosed workspace where some other desks are visible and dislike a workspace that is mostly enclosed. Furthermore, Office-workers preferred a large (13 m 2 ) workspace.
Concerning the crowdedness of the workspace, Office-workers preferred a workspace that is not busy nor calm where some desks are occupied or a workspace that is calm with lots of empty desks, but not a workspace that was busy where almost all surrounding desks are occupied. For the Home-workers, on the other hand, crowdedness was a very important aspect, they clearly preferred a calmer workspace than the Office-workers. Then, the preferences vary for concentration spaces. The Officeworkers preferred the availability of concentration spaces the most given the concentration-scenario, while the Home-workers had no clear preference. Office-workers also preferred a workspace with sufficient communication spaces.
The model output also indicates to what extent employees prefer a workspace in the office or to work from home. Table 5 shows the differences in probabilities that a workspace will be chosen per activityscenario. The Office-workers were most likely (99%) to choose a workspace in the office over working from home in the communicationscenario, while this is only 34% for the Home-workers segment. In the 50/50-scenario as well, the Office-workers were most likely (95%) to choose the workspace in the office. Again, for respondents in the Homeworkers class, this probability was only 16%. However, for the concentration-scenario, both segments preferred working from home over a workspace in the office (respectively 75% and 90%).

Segment profiles
The next step was to test whether the segments can be explained by personal, work-related, or home workspace-related characteristics. It was tested with Chi-square (Table 6) and T-tests (Table 7) whether differences between the two segments were found. Only significant findings are presented.

Personal-and demographic characteristics
The difference in gender between the two segments showed on average more males in the Office workers segment and relatively more females in the Home-workers segment. Also, there were more highly educated respondents among the Office-workers than among the Homeworkers. In terms of household type, there were relatively more single households in the Home-workers segment and more respondents that lived with other roommates in the Office-workers segment, and also more couples without children in the Office-workers than in the Homeworkers segment.

Work-related characteristics
The work-related variables indicated that the Office-workers segment included more employees with a job at Consultant/surveyorlevel than the Home-workers segment. This latter segment had more employees with a job on the administrative level. Moreover, all employees working at the (International) Partner level belonged to the Office-workers segment, while relatively more Head/Lead level employees could be found in the Home-workers segment. In general, Officeworkers worked more hours than the Home-workers. In terms of time spend, not surprisingly Office-workers spent more time in the office and Home-workers at home and they already did so before COVID-19. But the home workers also spent more time on individual concentrated work than the Office-workers. Travel time seemed to influence the preference for working in the office or at home as well. Respondents that spent less than 15 min of their time commuting from home to the office belonged more to the Office-workers, while respondents that spent more than 61 min commuting more often were a Home-worker.

Home workspace-related characteristics
The only difference found between the two segments for the home workspace shows that the Office-workers were more disturbed at home by noise than the Home-workers.

Discussion and implications for theory
Although the expected general tendency is to do concentration work at home and communicative work at the office in the future, this study shows that this is not the case for everybody. To the best of our knowledge, this is the first study showing stated choices where to work for specific work activity scenarios in relation to available office and home workspace designs. Such insights are valuable, not only because they confirm that the employee-workplace alignment mechanism is not a one-size-fits-all solution, but also because they show the need to keep a diverse office workplace available to satisfy needs of all employees if hybrid working policies are introduced. The current call in practice to focus office environments even more on communication activities would not do well for a significant part of this case organization's employees (the office-workers) and probably also not for many others with similar employees/tasks. On the contrary, the homeworkers group appears inclined to even do communicative work from home as well. As mentioned earlier, both pre-Covid studies (Derks, Agterberg, Beumer, & Weel, 2011;Peters et al., 2009) and studies during the pandemic (Waizenegger, McKenna, Cai, & Bendz, 2020) have shown that communication suffers from working from home. This calls for more research on what type of communication can be done from home, when employees are able to work hybrid again without lockdowns of other restrictions/fears related to Covid-infections. An additional topic for such studies should be to identify how hybrid working can be done without hurting social

wellbeing.
A novel finding of this study when regarding existing research on relevant workspace characteristics to support communication and concentration, is the relative importance of the different workspace characteristics in the decision where to work. Much of the scientific debate so far is focused on open versus closed/cellular offices (e.g. James, Delfabbro, & King, 2021;De Croon et al., 2005) and the detrimental effects of noise (e.g. Jahncke et al., 2011). While this study confirms the importance of noise perceptions, it also shows that the openness might not be the most important aspect to determine workspace choices. Besides noise, rather the crowdedness on the floor and position of the desk in relation to walking routes in the open office parts appear to be important. Even on days with lots of communication activities, people still did not prefer to hear intelligible conversations when they are sitting at their desk. This calls for a different way of thinking about activity-based office design research. As society will want to draw employees from the 'home-workers' segment to the office (sometimes) as wellto encourage the so desired and important unplanned communication for organizational success and creativityscientists will need to identify ways to reduce feelings of crowdedness and noise disturbance at the open spaces in offices. The frustration for employees of a workspace surrounded by regular intelligible conversations and/or a busy floor where almost all surrounding desks are occupied has been known for a long time (e.g Schlittmeier, Hellbrück, Thaden, & Vorländer, 2008;Sundstrom et al., 1994). Nonetheless, studies have hardly identified and tested potential solutions for what now seem to be the most important determinants in future hybrid working inclinations.
Another interesting addition to theory, is the clear aversion of sitting in an isolated spot (not near a walking route). Despite the previously identified presence of distractions in both traditional and activity-based offices (Budie et al., 2019;De Been et al., 2015), our findings show that isolation is not the way employees want to see this solved. Cobaleda--Cordero, Babapour, and Karlsson (2020) showed that other ways to isolate oneself from noise (quiet rooms) in activity-based offices are hardly used either. Together this seems to suggest a need for research to identify a limit for how large an open area can be without needing such isolated spots for concentration tasks. This would be a different approach in trying to deal with the "privacy-communication" trade-off mentioned by . On the other hand, the  R. Appel-Meulenbroek et al. findings also stress that both sufficient communication spaces and concentration spaces should be provided in the building, which contradicts with the study by Cobaleda-Cordero et al. (2020) that such concentration spaces are not used much. So, it merits to have further research on how to improve concentration spaces to optimally support hybrid working. Although a large proportion of respondents prefer to do their concentrative work at home, many of the office workers employee class prefer to do both concentrative and communicative work in the office. Maybe because of social distancing rules, a relatively large individual workspace size in the open area is preferred by these 'office-workers' as well. So, providing the office with enough and appropriate breakout rooms of different types and offering large individual workspaces could be a win-win situation, because interactions might take place less in the proximity of other workspaces which prevents possible distractions.
The two segments distinguished in this study also showed interesting differences in the personal characteristics that were identified from theory. Where the office workers segment contained relatively more male, highly educated, full-time workers with communication as an important component of their work and a short commute (under 15 min), the home workers segment contained relatively more females, parttime employees, and administrative functions, plus more individually focussed work and workers with a long commute (over 60 min). These findings intuitively seem to make sense, except perhaps the gender differences as there is much contradictory evidence from other recent studies. For example, Barrero, et al. (2021Barrero, et al. ( , pp. 2020Barrero, et al. ( -2174 showed in their study that after COVID-19, men want to work more from home than women (26% of working days for the male respondents versus 18% for the female respondents) even though women valued working from home as much as men did. In addition, another study showed that men were more satisfied with working from home than women because they agree more that working from home leads to a better combination between family and work (Timmers, Van Puyenbroeck, & Emmery, 2020). It appears that women, in general, feel more responsible for caregiving activities (childcare or care for a family member) and spent more time on household chores than men do (Del Boca, Oggero, Profeta, & Rossi, 2020;Donnelly & Procor-Thomson, 2015;Wheatley, 2012). In exceptional situations, like COVID-19, the extra care of children or family members that is already falling on women's shoulders further increased, due to reduced access to childcare, schooling, or eldercare facilities. Especially, working women with children aged 0-5 found the work-life balance during COVID-19 more difficult (Del Boca et al., 2020). An unexpected finding was therefore that couples with children were represented similarly in both groups, and couples without children were over-represented in the office workers segment. This does not seem to confirm that children at home are affecting the ability to work as much as suggested by Soetman (2011). OudeHengel, Bouwens, Zoomer, DeVroome, and Hooftman a Significance (p= < 0.10, *p < 0.05, **p < 0.01***). (2021) also showed that especially women were able to get more work done at home and older studies showed additional benefits for women while working from home (Peters et al., 2009). A last novelty of this study is that it showed how the home workspace determines to which segment one belongs. The Office workers more often experienced noise disturbance at home and/or lived with other people in the house. This could even be related to the findings on gender differences, as previous studies show that women, in general, have greater sensitivity for noise (McFadden, 1998;Velle, 1987) and thus it might play a bigger role in their decision. It appears that only women experience fear and physiologic arousal (e.g. raise in blood pressure, accelerated respiration) when they are faced with noise, while men react with surprise and only report subjective arousal (Rhudy & Meagher, 2001). But further research is necessary in this area as well. In addition, Kim, De Dear, Candido, Zhang, and Arens (2013) showed that female office workers are in general less satisfied with Indoor Environmental Quality factors at the office than their male counterparts, which could also be why they might prefer the comfort of their own home. Clearly, further research is necessary.

Limitations and future research
There were a couple of limitations in this study. It focussed on physical workspace characteristics that influence both concentrative and communicative work. However, there are many more aspects that affect either concentration (e.g. daylight, temperature, or personal circumstances) or communication (e.g. cultures, hierarchical organisation, or (unwritten) rules of conduct) or workspace preferences in general. In addition, this study distinguished only three activity-scenarios (communication-scenario, 50/50-scenario, and concentrationscenario). So, future research should investigate the concentration activities and communication activities more in detail for optimal support. Moreover, the sample size of this study was not considerably large (N = 218) and included only respondents from one organisation. Therefore, in further research, a larger sample would be recommended, and it would be interesting to find out whether there are differences in results between different types of organisations. Last, the addition of social and organisational context variables of the employees would be valuable, to identify how aspects such team motives and conflicting interests of organisations and their employees determine choice of work location.

Implications for practice
For Corporate Real Estate Managers (CREM), it is important to understand how to create an optimal work environment for employees to increase both individual outcomes (e.g. productivity, satisfaction, wellbeing, etc.) and the performance of the organisation as a whole (e.g. knowledge sharing, coordination, organisational culture). In addition, the quality of the workspace can be part of the talent strategy used to attract, retain, motivate, and engage skilled employees (Earle, 2003). CREM and human resource departments can use the information from this study to get more insight in which work activities their own employees prefer to do where, and which considerations the employees make when choosing to work in the office versus to work from home.
A strong aversion to intelligible office noise is again confirmed in this study, while especially these intelligible conversations have been shown to increase spontaneous knowledge sharing (Appel-Meulenbroek et al., 2017). So, organisations that care strongly about this, might also consider 'training' their employees in dealing better with noise disturbances. Research has shown that not even half of office workers speak up when facing unwanted noise and 21% does nothing to cope with it (Appel-Meulenbroek, Steps, Wenmaekers, & Arentze, 2021). Training could help to increase their understanding that they also benefit from such conversations, which might reduce frustrations. In general, it could be valuable to offer employees guidance on how to choose where to work; both regarding the office versus home choice and regarding the workspace type chosen inside an activity-based office. Each person "has flaws in their reasoning and uses heuristics, instinct, and emotions in decision-making instead of perfect logic to calculate probability for the best utility after exhausting all comparisons between pros and cons" (Lee, 2021, p. 209). Guidance on work location and workspace choices might increase workspace switching behavior when activities switch during the day and with that employee-workplace fit perceptions (Hoendervanger et al., 2016). Perhaps technologies, such as apps showing occupancy rates and availability of spaces can help as well.
Overall, this study contributes to a better understanding of the sometimes-contradictory office workspace requirements and preferences for concentrative and communicative work. It is important for organisations to make clear agreements with employees on hybrid working behavior to ensure a good balance between working from home and working at the office. This will also allow employees to benefit from both the positive effects of working from home (e.g. increased productivity, more flexibility) and those from working in the office (e.g. knowledge sharing, strengthening collaborative relationships).

Statements
• This research has been approved by the Ethical Review Board of the Eindhoven University of Technology under approval number ERB2020BE61.

Declaration of competing interest
We have no competing interests to declare.