In this paper, we investigated the role played by housing functions in the residential mobility of the tenants of a real estate owner and two of the largest cooperatives in Switzerland. Based on prior qualitative research, we introduced a multi-step theoretical model of tenants’ decision to relocate (Fig. 2) and then explored its linkages by means of empirical analyses of survey data.
In the following subsections, we discuss the results of this study along four lines: first, we present a synthesis of the findings and their theoretical contribution; second, we illustrate potential implications for practice; third, we discuss the study’s limitations; lastly, we identify promising avenues for future research.
Research findings in perspective: disentangling systems complexity
The first hypothesis scrutinized in this study was that housing functions can be used as proxies for residential attributes (housing, neighbourhood, location) to understand households’ satisfaction with their dwellings and thus are relevant for unravelling the decision to move and the selection process (H1).
Results have shown that, in most cases, residential satisfaction is more likely to increase with a decreasing gap between the housing functions of the ideal and current dwelling. However, we also observed that these findings are not consistently significant across categories of satisfaction. More specifically, the fulfilment of a housing function was found to make large or little-to-no difference to tenants’ residential satisfaction—e.g., for certain functions the gap was a significant predictor of the jump between a strong dissatisfaction to a strong satisfaction or vice-versa, for others of the jump between ‘neither nor’ to ‘strongly satisfied’ or vice-versa. In agreement with recent studies that have disproved the commonly explored existence of a linear relation between satisfaction and gap (see e.g. Jiang et al., 2020), our choice of a multinomial regression model pointed to the different influences that housing functions can have on rather than across categories of satisfaction. Also, our findings contribute to the research of the many scholars who, since the seminal work of Rossi (1955), have attempted to disentangle the complex links between residential satisfaction and the determinants of residential mobility (see, for instance, the conceptual model proposed by Marans, 1976). In particular, the existence of a direct or ‘mechanistic’ relationship between the residential environment and household satisfaction has often been questioned (Lawrence, 1987; Michelson, 1980), arguing that the latter can vary within and between households who subjectively interpret and assess the objective characteristics of the former (i.e. (dis)amenities), depending on a variety of factors (expectations, reference groups, subjective beliefs; Cook & Bruin, 1994; Diaz-Serrano & Stoyanova, 2010; Galster, 1987; Galster & Hesser, 1981; Jansen, 2014; Jiang et al., 2020; Marans, 1976). By introducing the functions as mediators between the human and material subsystems and thereby accounting for both tenants’ preferences and dwelling forms, this study does not advocate for the existence of a direct relationship between satisfaction and dwelling but rather an indirect and systemic one. This conceptualization makes it possible to overcome the limitations encountered in other authors’ empirical analyses, and in particular the aforementioned subjective ways but also the complex combinations in which dwellings features affect residential satisfaction—i.e. the correlations between and within categories of residential attributes (dwelling, neighbourhood, location) or the different effects that each of these categories has been found to exert on residential satisfaction (Jiang et al., 2017; Molin et al., 1996; Wong, 2002). In other words, our results demonstrated that the notion of housing function can offer a shortcut to link residential satisfaction to the objective and subjective characteristics of the environment and of its residents while accounting for the system’s complexity.
The findings of H1 are of relevance given that residential satisfaction plays a role in the decision to move and the formulation of preferences for the new dwelling. When looking at the former, we found that housing functions both directly and indirectly influence the extent to which tenants are likely to move following an event (e.g. a new child; H2). More specifically, we observed that the level of residential satisfaction (which itself is influenced by the size of the gap between ideal and current functions) and the function that the dwelling fulfils are significant explanatory variables of the event triggering the move. Building on the seminal work of Speare (1974), most scholars have examined the direct and indirect relations between households’ mobility, residential satisfaction, housing features and socio-demographic characteristics (for an overview, see Jiang et al., 2017). Our findings contribute to this body of literature by focusing on the effects that satisfaction and housing functions (and therefore housing and residents’ characteristics) have on the triggers of the relocation process, rather than on the intention and actual behaviour.
Similar research was undertaken by Wong (2002), whose results showed that the triggers to move have ‘unequal correlations’ with households’ level of satisfaction (p. 227). By grouping triggers into types (i.e. opportunities, problems to solve, and radical changes), our model extends her results one step further. More specifically, when comparing Table 4 with Table 5 and confirming former exploratory findings (Pagani & Binder, 2021), we observe that the trigger events that are the most effective with an increasing level of satisfaction are often the predictors of the imposed triggers or ‘forced’ moves (Clark & Onaka, 1983), i.e. ‘radical change’ and ‘problem-solving’.
When looking at the formulation of residential preferences and by further exploring the systems interrelations between housing functions and triggers, our findings demonstrate that trigger types differently arbitrate the change in function for the new dwelling (H3). More specifically, despite the weak-to-medium effect size, a radical change was found to most strongly affect tenants’ preferences in terms of housing functions. This finding first supports the argument that relocations are instrumental to goals, which can change during the household’s life course (Coolen et al., 2002; Mulder & Hooimeijer, 1999); second, it corroborates H1 by showing that housing functions are a constituent element of these ‘goals’.
Based on this observation and on the body of literature introduced in this paper, households’ characteristics were expected to influence housing functions in multiple ways (H4). Our regression models confirm that household type (marital status, age and children) is a significant explanatory variable for five of the nine ideal housing functions. The findings also illustrate the diversity of ideal dwellings resulting from combinations of different careers (e.g. educational, occupational; Mulder & Hooimeijer, 1999), including the type of tenancy.
As outlined in this section, our findings contribute to the body of literature on residential mobility by illustrating the potential of introducing the notion of housing functions for disentangling the complexity of the human–environment system under study. More specifically, our results suggest that the functions orchestrate the factors leading to the moves of Swiss tenants (i.e. triggers, satisfaction and preferences).
Relevance for practice
In agreement with several scholars, this study argued that a better understanding of the relocation process and its determinants can play a key role in fostering the provision of adequate, appropriate, and quality housing—i.e. dwellings that support and meet the culture, values and needs of households for which those are intended (see for instance Clark et al., 2006; Franklin, 2001; Kahlmeier et al., 2001; Lawrence, 2004; Molin et al., 1996; Rapoport, 1977). Due to the housing system’s complexity, disagreement between housing providers (i.e. owners, practitioners, policy makers) and users (i.e. residents) on what constitutes residential quality persists (Diaz-Serrano & Stoyanova, 2010; Franklin, 2001; Jansen, 2014; Lawrence, 2009, 2021a; Marans, 1976), which can have several implications. For instance, the difficulty in understanding the links between objective and subjective assessments of the residential environment can undermine the success of housing developments or neighbourhoods—when the housing situation is dissatisfactory, the residents consider housing alternatives (Cook & Bruin, 1994; Kwon & Beamish, 2013; Lawrence, 2009); also, dissatisfaction has been demonstrated to have repercussions beyond households’ relocation, and especially to impact residents’ health and well-being (Clark & Kearns, 2012; Jansen, 2014; Kahlmeier et al., 2001; Rolfe et al., 2020).
For these reasons, it has long been argued that plans and programs related to providing or improving housing quality must include final users in the discussion (Lawrence, 2021a). However, participatory approaches might be insufficient if tools to disentangle the system’s complexity and foster the integration of the multiple stakeholders’ perspectives are not available. Therefore, based on the results presented in our study, practitioners should consider the added value of adopting a systems perspective and using the notion of housing functions for accounting for the relative value that different residents’ groups attach to specific dwelling, neighbourhood and location features while ensuring a comprehensive assessment and provision of the many ‘interrelated purposes that impinge upon the quality of the [residential] environment’ (Lawrence, 1995, p. 1663).
While the multi-step model proposed in this study offers a new take on the conceptualization of the residential mobility process, several limitations must be acknowledged. Mainly, the results of the analyses were not consistently significant for the nine housing functions: on the one hand, they were sensitive to the chosen regression models (i.e. ordinal, multinomial; e.g. Table 14); on the other hand, they were influenced by the choice of the variable to investigate. Below, we discuss the effects of models and variables on our results.
Gap and satisfaction
Looking at the data of Table 3, four of the nine functions are not significant in the regression model. When comparing it with Table 8, it can be observed that ‘commodity’, ‘impermanence’ and ‘security’ are on average fulfilled more than tenants desire (see variable ΔCurrent-Ideal). This result shows the limitation of the formula chosen to compute the gap between reality and preferences, which considers only the lack of a dwelling function as a predictor of residential satisfaction, regardless of its abundance. Rather, more complex models have assumed the existence of an ideal point, whereby satisfaction decreases if reality deviates from aspirations in both directions (see e.g. Jiang et al., 2017, 2020); in other words, a function might also be perceived as undesirable or conflictual and thereby negatively affect tenants’ level of satisfaction (e.g. a dwelling ‘free from tradition and memory’ versus the need for a place ‘where I feel rooted’). In addition, to account for residents’ different sensitivities to under- and outperformance of a preference, Jiang and colleagues (2020) proposed non-linear asymmetric gap models which consider that the same gap might not always lead to the same level of dissatisfaction. Also, beside the generally used difference formulation, the authors computed the size of the gap as a relative difference, i.e. dependent on how great the level of aspiration is.
Aside from the way variables were computed, the predominance of moderately and totally satisfied tenants in the sample of respondents is a relevant limitation (see Table 10); this bias or dissonance is common in other studies, and derive from a tendency of evaluating a past decision positively (Jansen, 2014; Kahlmeier et al., 2001; Marans, 1976).
In sum, residential satisfaction is a complex notion that has been conceptualized, measured, and calculated in manifold ways and is subjected to several biases. In this study, the way the dependent and independent variables were computed revealed several limitations which could be overcome by more methodologically advanced gap models.
Asking tenants to assign the trigger events to one of the three proposed types aimed at validating the typology of triggers proposed in the Pagani and Binder’s (2021) qualitative study. However, while observing the richness of events that can be categorized as problems to solve or radical changes, we also faced the issue of having the same event categorized in both types.
More specifically, a closer examination of Tables 5 and 11 shows that the links between functions, trigger events and trigger types remain unclear. For instance, the function ‘property’ was found on the one hand to increase the likelihood of moving due to trigger events categorized as ‘radical changes’ or ‘problems to solve’ and on the other hand to decrease the likelihood of moving due to a ‘decrease in comfort’, which tenants also classified as a problem to solve. Another example is Table 6, where an update in housing functions—which was only expected for the category ‘radical change’—was observed following the trigger ‘problem-solving’, a result that could also be explained by the above-mentioned overlapping of types per event.
These unclear relationships potentially suggest the existence of sub-categories of the three trigger types depending on the triggering ‘power’ of each event in the type, meaning the level of satisfaction at which they are effective.
ΔFunctions and trigger types
The choice to compare changes in current housing functions (i.e. between past and present dwelling) to observe the effects of triggers on residential preferences should also be discussed. One could argue that this approach is correct only if the current housing function (i.e. revealed preferences) corresponds to the ideal one (i.e. stated preferences). If not, the tenant would take advantage of any trigger type to choose a dwelling that better matches its ideal functions (Pagani & Binder, 2021). At time t, the result of the move would evince an update in current functions, which would not correspond to an update in the ideal ones.
In agreement with this argument, results for the subsample who moved with a high level of satisfaction (i.e. with current and ideal functions matching; see H1) showed improved results compared with the full sample (Table 6). However, contrary to H3, the trigger ‘problem-solving’ brought about an unexpected and significantly greater change than the trigger ‘opportunity’. Possible explanations for this result emerge when considering the context, as illustrated in the next subsection.
Beyond variables: the relevance of the context
The extraordinarily low vacancy rate in Switzerland cannot be overlooked when investigating tenants’ residential choices. Although encompassed by the trigger events, the influence of micro- and macro-level contexts was not thoroughly accounted for in the variables chosen for our analysis of preferences. In fact, analysing the stated and revealed preferences through ideal and current housing functions did not account for the adjustments of the criteria to what is possible (Timmermans et al., 1994; van Ham, 2012); elements such as income or the availability of dwellings on the market can make preferences and final selections deviate from ideal housing functions. This is clear in Sect. 4.5, where salary and education were found in most cases not to be good predictors of ideal housing functions. This argument is also key for our interpretation of Sect. 4.4, whereby the trigger ‘problem-solving’ was found to bring about an unexpected change in function; considering time constraints (i.e. contract expiration), a compromise between the dwellings available on the market and the ideal one is often needed, thereby potentially resulting in a change in function. Further, the results presented in Sect. 4.1 show that fulfilment of the function ‘production, consumption’—which encompasses basic activities such as laundering or social activities such as companionship—is relevant but not sufficiently critical to discriminate a low from a high level of satisfaction; this finding should be further investigated in relation to the Swiss economic and sociocultural context (e.g. wealth, interpersonal relationships).
Previous studies have accounted for resources and restrictions (e.g. household salary), and opportunities and constraints (e.g. vacant dwellings) when investigating the decision process by adopting the so-called ‘three-stages approach’ (Mulder, 1996; Mulder & Hooimeijer, 1999). Following this approach, a new function could be introduced: the desired function. As the ideal function is only dependent on a household’s trajectories, the desired function would correspond to the adaptation of the ideal one to resources and restrictions, and the current function to the adaptation of the desired one to opportunities and constraints. These three types of functions would more specifically account for the trade-off between the multiple determinants that arise from, for example, lifestyle and individual resources (Thomas & Pattaroni, 2012) and the re-evaluation of preferences in the search process (Brown & Moore, 1970).
Based on the limitations illustrated above, it becomes clear that further research is needed. Firstly, the role of housing functions in the selection process should be more closely considered by (1) focusing on the readjustment of the ideal housing function(s) to the desired one(s) following a trigger and of the latter to the current one(s) for the final selection; (2) critically analysing the contribution of the three types of functions to households’ satisfaction with and selection of a dwelling; and (3) exploring the potential to use previously-identified explanatory variables for tenants such as age, size of household and rent as predictors of the desired function (Clark & Dieleman, 1996). In particular, further studies of the relationship between housing functions and resident satisfaction could benefit from the substantial methodological advances in the field, e.g. the use of non-linear models (Jiang et al., 2020).
Secondly, while our study investigated tenants’ past move—where the intention to move corresponds to actual residential mobility—new insights could be gained by examining unsuccessful relocations (Coulter, 2013); in this context, the factors preventing relocation identified in the large amount of research based on the stress-resistance models could be explored in relationship to housing function and trigger types (i.e. the monetary and non-monetary costs of moving; see Brown & Moore, 1970; Clark & Onaka, 1983; Goodman, 1976; Mulder, 1996; Phipps, 1989; Phipps & Carter, 1978; Wolpert, 1965).
Thirdly, this paper presented the results of quantitative research conducted in the framework of the Swiss rental market which are country- and tenure-specific; considering the relevance of the context for the present and future studies, the tenancy type and the influence it has on tenants’ decisions could also benefit from further research (e.g. due to occupancy rules, a reduction in household size can result in a ‘forced move’ for cooperative tenants). Furthermore, while the notion of housing functions allowed us to consider and have a better understanding of the interrelationships at play in the housing system (i.e. objective and subjective assessments of housing quality, changes in residential preferences, residential satisfaction, etc.), additional qualitative and quantitative research could be conducted to explore the functions’ potential material manifestations in the Swiss context for different inhabitants’ groups.
Lastly, for our results to appeal to decision-makers and practitioners, and thereby reduce the so-called ‘applicability gap’ (Lawrence, 2021b), the proposed model of residential mobility should be explicitly integrated with context dynamics, i.e. opportunities and constraints generated by the housing market. Since a systems perspective was adopted, an agent-based model (ABM) can be utilized for this purpose. The goal of an ABM is to observe the parallel actions of components and their interaction, thereby discovering emergent properties from a bottom-up perspective (Nikolic & Ghorbani, 2011). Implementing an ABM would make it possible to simulate the system outlined in this paper (i.e. tenants’ residential relocation process) and integrate it with housing stock dynamics (i.e. construction, demolition, renovation). By accounting for the material components of housing and stakeholders’ goals, priorities and values, the model would contribute to a greater understanding of the behaviour of such a complex human–environment system and thereby make it possible to observe otherwise-unpredictable reciprocal effects between residential preferences and dwellings.