Merit, Need, Entitlement? Investigating Fairness of Housing Evaluations

1 Introduction

Empirical social justice research investigates (among other issues) individuals’ fairness perceptions regarding the unequal distribution of scarce valuable goods in societal aggregates. The literature on distributive justice has shown that people have normative ideas about “who deserves what” (what amount of goods or rewards for whom?; Aalberg 2003: 24) and apply the justice principles of equality, merit/equity, need, and status entitlement when making fairness evaluations (Adriaans & Fourré 2022; Reeskens & van Oorschot 2013). Equality refers to the ideal that everyone should receive the same amount of rewards or resources. According to the equity or merit principle, resources should be allocated on the basis of individual commitment, performance, or contribution: those who perform better or contribute more deserve higher amounts of rewards. The need principle states that resources should be allocated according to personal need (e. g., having children, unexpectedly becoming unemployed, having health problems). Status entitlement refers to the principle that ascribed or achieved social status characteristics, such as gender, ethnic origin, or occupational status, are taken into account when making distributive justice evaluations: some individuals deserve more because of their anticipated positioning on the social ladder (Adriaans & Fourré 2022; Arts & Gelissen 2001; Hülle et al. 2018). Empirical research – to a large degree relying on factorial survey (FS) experiments – has investigated the role of these justice principles in regard to many topics and inequality dimensions, such as earnings or wages (e. g., Alves & Rossi 1978; Auspurg et al. 2017; Shamon & Dülmer 2014) and environmental goods or burdens (Liebe & Dobers 2020), and also for policy issues like taxation (Gross et al. 2017; Groß & Lang 2018; Sachweh & Eicher 2023), unemployment benefits and pension schemes (Reeskens & van Oorschot 2013), and childcare fees (Busemeyer & Goerres 2020).

However, up to now, fairness evaluations of unequal residential conditions have been neglected by existing research. How do people perceive different housing circumstances (for example, living space, housing costs, building type or construction quality, and neighborhood quality) with respect to the above-mentioned justice principles? Who deserves what in terms of housing? Is it, for instance, individual – for example, education-, job-, or income-related – performance that should decide about more favorable housing conditions? Or is the need principle – e. g., having children or facing unforeseen financial troubles – more important when individuals evaluate the fairness of housing conditions?^[1] And do people also revert to status characteristics or group membership when evaluating these questions in the sense that one social (status) group is more entitled than another?

As housing disparities represent an important social inequality dimension, it is astonishing that this topic has not yet been picked up by empirical justice research. Unequal housing or residential conditions can refer to the dwelling, home, or house and its various characteristics (e. g., living space per capita, housing costs, homeownership), but also encompass inequality-relevant aspects of the living environment, such as neighborhood characteristics (e. g., crime rate, availability of green space), infrastructure, or environmental burdens. Many studies have confirmed pronounced inequality in regard to various indicators of residential conditions (see for example Diekmann et al. 2022; Holm et al. 2021; Kohl et al. 2019; Kolb 2013; Rüttenauer 2018; Rüttenauer 2019; Rüttenauer & Best 2021). Further, housing inequality is strongly related to other dimensions of inequality, such as income and social status. In this respect, inequalities can be said to accumulate: people who are relatively worse off in the latter dimensions are also worse off in terms of their housing conditions (Filandri & Olagnero 2014; Hinz & Auspurg 2017; Kurz 2000, 2001). Also, discriminatory mechanisms are at play in the access to housing (Auspurg et al. 2019; Horr et al. 2018), and people also seem to have discriminatory or segregational residential preferences concerning neighborhood composition (Lewis et al. 2011; Wolter et al. 2023; Zangger 2021). Finally, residential inequality has been on the rise in recent years (e. g., Burrows & Knowles 2019; Butterwegge 2021; Dewilde & De Decker 2016; Dorling et al. 2005; Helbig & Jähnen 2018; Wind et al. 2017).

Given the lack of research on fairness of housing evaluations, this article aims to investigate how people assess the fact that different people live in different housing conditions with respect to fairness and distributive justice principles. To what extent are the criteria of merit/equity, need, and status entitlement decisive when it comes to the question of which housing conditions are perceived as fair or just? Firstly, I argue that individuals generally apply these principles when making fairness of housing evaluations. There are plausible reasons to assume that the criteria of merit, need, and status entitlement are taken into account and exert effects that are comparable to those found for other inequality dimensions, notably in the fairness of earnings literature but also in regard to other issues. This includes, with respect to status entitlement, examining whether gender and occupational status play a role: are women and lower-status occupations entitled to fewer rewards, as found in the fairness of earnings literature? Secondly, regarding the question of whether merit/equity or need is more important, I introduce two competing hypotheses put forward by Reeskens & van Oorschot (2013), who distinguish self-inflicted, predictable risks (for which the merit principle should dominate) from exogenous, unpredictable risks (for which the need principle is dominant), and I discuss the implications for the issue of housing disparities. Thirdly, by extending the status entitlement argument, I also ask whether “double standards” (Foschi 1996, 2000; Jasso & Webster 1997) are at play, i. e., whether merit, need, and other criteria are applied differently depending on gender and occupation. In this regard, the hypothesis is that harsher standards are applied to disadvantaged groups – in our case, women and those with lower-status occupations. Apart from these main content-related questions, the study also aims to generate insights on the methodological side by demonstrating how FS experiments – the state-of-the-art method for investigating justice evaluations – can be applied to the topic of the fairness of housing. This includes discussing the limitations of the methodological setup used and resulting research desiderata.

To address these issues, I present findings from a vignette study carried out in fall 2022, as part of the Konstanz Citizen Survey (“Konstanzer Bürgerbefragung”; N = 1,154) conducted in the City of Konstanz (Germany). Konstanz is a mid-sized university city (counting approximately 87,000 inhabitants) in the far south of Germany, with a tight housing market and comparatively high rents.^[2] It is therefore typical of many German cities and agglomerations with similarly competitive housing markets. Respondents rated vignettes containing example residential situations of fictitious persons with respect to whether the described housing constellations are “fair,” “unfairly too bad,” or “unfairly too good.” The vignette dimensions varied (i) inequality-relevant housing characteristics (living space, costs, location), (ii) job performance of the fictitious person (merit), (iii) need factors (having children, living with a partner), and (iv) gender and occupation as factors prone to possible status entitlement effects and double standards.

The article seeks to contribute both to empirical justice research – a well-established field, but one which thus far has been silent about residential inequality – and to the ongoing public and scientific discussion about (rising) residential inequality (often referred to by the catchphrase “housing – the new social issue?”). To the best of my knowledge, the work presented here is the first to use an FS approach to examine fairness of housing evaluations.

The article proceeds as follows. Section 2 sums up the core concepts and findings from the theoretical and empirical justice research literature and lays out the main hypotheses regarding distributive justice principles and the double standards hypothesis. Section 3 presents the research strategy, study design, and methods. The results are reported in Section 4. The article concludes in Section 5 with a discussion and, as it proposes a new research field, desiderata and suggestions regarding future research.

2 Empirical Justice Research: Concepts, Theory, and State of Research

3 Research Strategy, Study Design, and Methods

This article aims to investigate how the above-presented justice principles and theoretical arguments apply to judgments about “who deserves what” in terms of housing (inequality). The research strategy draws on the method of FS experiments together with the “multiple standard framework” which relates to the “justice evaluation function” proposed by Jasso (1980, 1996; for a summary, see Auspurg et al. 2017). This framework sets into relationship a (1) justice evaluation by survey respondents, given (2) an actual allocation or distribution scheme, and (3) their notion of what amount of reward or allocation would be fair. The former two are given by the dependent variable and example constellations of the vignette setup (see below). In this way, it is possible to estimate the third component, i. e., what amount of reward respondents regard as fair. This approach thus requires the FS experiment to include indicators for the justice principles (merit, need, entitlement) and the reward or allocation scheme: in our case, indicators for residential conditions or residential inequality. Further, the dependent variable has to measure what respondents regard as “fair” or “unfair” in as detailed a way as possible.

The main advantage of using FS experiments to investigate justice evaluations consists in combining a high internal validity – the ability to establish causal effects by sticking to a strict experimental setup – with a high external validity, ensured by implementing them in large-scale population surveys (Auspurg & Hinz 2015; Treischl & Wolbring 2022). Further, respondents always have to trade off different aspects of the issue under consideration, which reduces ex-post rationalizations and social desirability effects and encourages them to avoid saying “yes” to everything when asked directly. In our case, respondents have to weigh the different justice principles against each other. One can expect that this will yield more valid assessments of their relative role for justice evaluations than if asked about them directly (“Do you think that high-performing people deserve better housing conditions than low-performing ones?”).

The study uses data from the fall 2022 wave of the Konstanz Citizen Survey, a panel study with at least one survey wave per year, carried out since 2008 by the Sociology Department of the University of Konstanz and the Konstanz City Council. The target population are all Konstanz residents aged older than 16 years having their first residence there. Apart from core modules that are surveyed each year, the questionnaire covers different topics in each wave. A key feature of the survey is that it contains topics both on communal issues that are relevant for the Konstanz public and for its political administration, but also scientific modules that are relevant for social science research. The sample is an offline recruited random sample from the official local population register; the response rate amounts to 55 percent, resulting in an analysis sample size of N = 1,154 respondents (see Spanner et al. 2023 for more details regarding response rates and further features of the survey). To adjust for selective response rates, all analyses are weighted with respect to the actual distribution of the Konstanz population regarding gender, age, nationality, and city borough. Table A1 in the online appendix presents descriptive statistics on some core socio-demographic variables. The questionnaire for the fall 2022 wave was multi-thematic. The module analyzed here was entitled “Justice of Housing” and contained the FS experiment, after which several questions were asked regarding the actual housing/living conditions of respondents.

The overall aim in setting up the FS experiment was to base the design on the setups in the fairness of earnings literature and related empirical justice literature. One core difference when it comes to housing inequality, however, is the fact that it entails not just one inequality dimension (earnings or wages or amount of income tax), but multiple dimensions: living space, homeownership status, housing costs, neighborhood characteristics, and many other aspects. In addition to these housing inequality indicators, vignette dimensions for the distributive justice principles of merit, need, and entitlement also had to be part of the FS design. Thus, the question was how to include as many of these aspects as possible in an FS setup by implementing a trade off with respect to restrictions regarding sample size and statistical power, i. e., the limited number of vignette dimensions that are feasible for one setup and given sample size. Further, there was a question as to whether households or individual persons should be chosen as the focus actors for the vignettes. Also discussed was the question of whether, in general, conducting a survey in a local – versus nationwide – context is better suited for studying fairness of housing evaluations. For this study, however, it was clear from the beginning of the project that the study would be implemented as part of the Konstanz Citizen Survey, and hence in a local context. An advantage of this is that one can assume that in a regionally limited context, there is less variation in the conditions of the local housing market than there is on a national level (concerning housing costs, for instance), and that respondents are better informed about local circumstances than about housing market conditions in geographically remote areas. Tailoring the vignette dimensions and levels to a local market, together with a more valid notion of what is realistic or not on the side of the respondents, should generate more valid and realistic fairness evaluations than if the same vignette setup were used for a nationwide survey. I will return to this point in the discussion section.

The vignette design was as follows. Respondents were introduced to the topic by informing them that the ensuing questionnaire section was about which residential conditions in Konstanz are perceived as just by the citizens (see the online appendix for the exact wording). Each respondent received six fictitious example constellations of a fictional person living in a certain residential arrangement. Due to complexity restrictions regarding the vignette setup, several factors were fixed by design. Respondents were told that they should imagine all example persons as being German, renting (i. e., not homeowners) an apartment (i. e., no detached or semi-detached houses or other dwelling types) in Konstanz. In the vignette texts, age was also fixed: all example persons were “in their mid-forties.” Table 1 depicts an example vignette, and Table 2 depicts the whole vignette universe.

The vignette universe is a 2⁴ × 3¹ × 4² × 5¹ design, which adds up to 3,840 vignettes. From this universe, a D-efficient sample was drawn according to the recommendations of Kuhfeld (2010), resulting in 264 vignettes, blocked into 44 decks. The vignette sample was optimized with respect to zero correlations between the vignette dimensions and all their second-order interactions; D-efficiency amounts to D = 96.2. No (potentially) implausible vignette combinations (such as floorspace of 1,400 m² together with only 500 € monthly costs) were excluded. The order of the six vignettes each respondent received was randomized. The choice of the apartment size (living space) and monthly costs was made on a realistic basis by taking the actual distributions of these variables in Konstanz into account.^[6]

The resulting vignette data comprise N = 6,924 vignette cases (see Tables A2–A4 in the online appendix for design-related statistics of the vignette setup). Due to item nonresponse, however, some cases drop out of the analysis. The data analysis employs standard multilevel regression methods (Hox 2010) and proceeds in three steps. First, a baseline model presents the main effects of the vignette variables on justice evaluations. This makes it possible to establish the absolute and relative importance of the vignette-level covariates for merit, need, entitlement, and residential characteristics. A second step then presents an illustrative-descriptive excursus: using calculations derived from estimates on cross-elasticities (more on this below) makes it possible to look at how the effects of the vignette variables relate to each other in terms of concrete figures on actual levels of living conditions for certain constellations (living space and monthly costs) that are perceived as fair (e. g., for having children versus no children, or for a medical doctor versus a factory worker). The third step of analysis focuses on the double standards hypothesis. This is done by estimating interaction effects between gender and occupation and merit, need, and inequality indicators in order to establish whether their effects vary depending on gender and occupational status.

4 Results: Fairness of Housing Evaluations

The dependent variable has a mean of 0.16 on the 11-point answer scale (median = 0), with a standard deviation of 2.5 points (see Figure A1 in the online appendix for a distributional graph). Nearly 30 percent of vignette answers are on the middle or “zero” category “fair/just”, but this is a typical pattern in FS studies (see, for instance, Lang & Groß 2020). In any case, the dependent variable has enough variance and respondents used all answer categories.

Figure 1 depicts the main results of the study by showing the effects of the vignette dimensions on respondents’ fairness of housing evaluations. Positive coefficients (above zero) mean that the respective vignette level is evaluated as being in the direction of “unfairly too good” residential conditions; negative ones operate in the direction of “unfairly too bad.”

Figure 1:

Main Effects of Vignette Variables on Justice Evaluation

Note: Linear multilevel regression, dependent variable: fairness evaluation of example (vignette) residential situation. Unstandardized regression coefficients and 95 %-CI (robust standard errors). See Table A5 in the online appendix for a full regression table with more information. N = 1,097 respondents and N = 6,529 vignettes. R² (McFadden) = 0.116.

The gender variable has no effect: male and female persons are evaluated equally with respect to which housing conditions are considered fair for them. In contrast, all levels of occupational status (compared to the reference category “nursing staff”) do have effects, and they are very interesting because they partly go against what has been hypothesized above. In particular, being a medical doctor or a manager goes along with being evaluated as having “unfairly too good” housing conditions. Hence, there are no status entitlement effects of the kind that accord those with higher-status occupations more favorable housing settings. Remember that such effects with respect to earnings/income have been found in various studies on the fairness of earnings (e. g., Alves & Rossi 1978; Auspurg et al. 2017, Hermkens & Boerman 1989; Sauer et al. 2009). Concerning housing, however, we see the exact opposite: a nurse and a factory worker are considered to be entitled to better housing conditions than a physician or manager. This finding, however, has to be relativized for the following reason: occupational status could have worked as a proxy variable for the vignette person’s income or financial situation, which was not explicitly varied in the vignettes. As the rent is held constant for different occupations, a manager is presumably assumed to have more money available than a nurse, and for this reason is evaluated in the direction of “too good.”^[7] The “currently unemployed” effect, in turn, has the expected direction: being unemployed is related to being evaluated as having “unfairly too good” housing conditions. The performance or merit indicator has the expected effect as well: if job (or job search) performance is low or unknown, the example housing situations are evaluated as being “unfairly too good” (the effect size for “unknown” being smaller) as compared to a high job performance. The effects of the need variables are also as presumed: living with a partner or with children goes along with being evaluated as deserving better housing circumstances. Finally, the indicators for living space and monthly costs have strong effects on justice evaluations, while the one for location is substantially lower. The living space effect is nonlinear: changes from large to very large apartment sizes have smaller marginal effects than changes from small to medium-large ones.

Regarding the distributive justice principles of merit, need, and entitlement, one can conclude that they all matter for fairness of housing evaluations. Comparing the effect sizes and asking the question “What matters more?” reveals that living space and monthly costs exert the largest effects (up to 3.3 and 2.4 scale points on the 11-point answer scale). The remaining effects are all within a range of up to one scale point, and hence are substantially much less important. Here, it is interesting that the equity principle, operationalized by job performance, does not dominate; the need principle (having children) has more impact. This finding is in favor of the second interpretation regarding the two competing hypotheses introduced above (self-inflicted, projectable versus exogenous, non-projectable).

One further result merits discussion: fitting a basic random-intercept-only model to the data (not documented) yields an intra-class correlation that is practically zero (ICC = 0.01). This means that, technically, no residual variance exists at the respondent level: the whole variance of vignette judgments (the dependent variable) is explained by vignette-level variables. There are two ways of interpreting this. First, it could be an artifact if, for example, respondents gave random answers due to a vignette design being too complex. Second, the zero ICC could be taken as an important finding in terms of content – namely, that respondents more or less completely agree in their judgments of the vignettes. This also entails that they would agree regarding the importance they attribute to the indicators for merit, need, entitlement, and housing conditions in their evaluations. If this interpretation holds, the finding is of utmost interest, as it shows that there are apparently no societal cleavages or tendencies toward polarization when it comes to the question of which housing conditions are just – at least on the local level that is focused on in this study.^[8]

The data give arguments in support of both interpretations. After entering the main effects of the vignette variables into the estimation, and hence controlling for the vignettes respondents receive, the vignette-level residual variance goes down by about one-half and the respondent-level error variance goes slightly up, resulting in an ICC of 14 percent. This would indicate that respondent-level factors do play a role for justice evaluations. However, there are two arguments as to why this should not be over-interpreted. Firstly, even the 14 percent respondent-level residual variance is still small when related to the whole residual variance in the empty model – the basic vignette effects model already reduces this variance by one-half. Secondly, I have conducted various robustness and exploratory analyses, checking for respondent-level effects (main effects of respondent-level variables and random slopes for the vignette variable effects). The checks for random slopes revealed that only being unemployed and having a very large living space had significant random slopes, but were substantially small. So, taken together, the effects of the vignette variables do not vary by respondents, especially those on merit and need. This also means that it is pointless to look for cross-level interactions. Concerning respondent-level main effects, I have checked for various ones, including homeownership, available living space, costs, social status, gender, having children, duration of stay in Konstanz, and city borough, which all had no effects on justice evaluations.^[9] The only factor that turned out to have an effect was age (older respondents tended to evaluate the vignettes as more “unfairly too good”, compared to younger ones). Further robustness checks also revealed that the zero ICC remains if all cases with a “0” answer (middle of the response scale) are dropped from the analysis.^[10]

Taken together, the data seem to go more in the direction of the second interpretation, i. e., that respondents do indeed agree in their fairness of housing evaluations. One could conjecture that this is owing to the local context in which this study was conducted, and that more disagreement would show up in geographically and socially broader contexts. And, of course, choosing a different vignette setup (in regard to dimensions and levels; and the introductory framework limiting the example constellations to rental apartments, etc.) could also have yielded different results in this regard. However, a consequence for the remainder of the article and for the study as a whole is that it is pointless to investigate respondent-level effects such as asking whether “haves or have-nots” (Reeskens & van Oorschot 2013) or politically left-and right-wing proponents have different notions of what housing conditions are fair, or whether certain groups of people accord less or more importance to the justice principles (heterogeneous treatment effects): there simply is no variance in the data that could be explained by such analyses.

How do the estimates of the vignette variable effects translate into concrete figures about which housing conditions are evaluated as being fair? This is addressed by an illustrative excursus in the next analysis step by relating the effects of the vignette variables to each other. This works for metric vignette variables – in our case, living space in square meters and monthly rental costs in Euros. For example, which living space is considered “fair” for a household with children, compared to one without? What rent level per square meter surface is considered fair? The methodological procedure used to answer these questions is analogous to calculating willingness-to-pay (WTP) estimates in choice experiments, or “just gender pay ratios” in fairness of earnings studies (cf. Auspurg & Hinz 2015: 100). For instance, the above-mentioned “just living space ratio” (JLSR) measured in square meters for having two versus no children is given by:

 (1)

where ß_children is the regression coefficient for having two children (versus not) and ß_livingspace is the coefficient for living space. For these models, the vignette variables “living space” and “monthly costs” were entered as metric covariates (documented in Tables A6–A8 in the online appendix).^[11] Further, I estimate an absolute and a relative ratio for each constellation: the first uses the original variables in units of Euros or square meters, and the second uses their natural logarithm, yielding relative ratios in percent units (Formula 2). For the derivation of the formulae, see Auspurg & Hinz (2015: 100) and Auspurg et al. (2017: online supplement).

 (2)

Table 3 shows the results. The price that Konstanz citizens regard as constituting a fair rent level amounts to approximately 14 € per square meter. Remember that this only refers to rented apartments, as the vignettes excluded homeowners and dwelling types other than apartments, notably detached houses. Most interesting here is that this “fair rent level” corresponds almost exactly to the actual price level for rented dwellings in Konstanz, which in 2020 was 13.6 € per square meter (median value, excluding flat shares but including detached houses: see Spanner et al. 2021: 31). With this level of rental prices, and as mentioned in the introduction, Konstanz is one of the most expensive cities in Germany. Despite this fact, however, respondents seem more or less to accept this and regard the rent level as fair. But it should be noted that this finding is subject to discussion. One could object, for instance, that this result naturally depends on the vignette dimensions and levels that respondents are responding to, and hence does not necessarily reflect “what really goes on.” In this regard, however, bear in mind that the ranges of rental costs and living space for the vignettes were chosen in a way that roughly corresponds to the actual ranges of these variables. But this in turn could represent another source of artifact: if respondents evaluated the vignettes with their known local market conditions as a reference frame in mind – mirrored by the ranges of the vignette dimensions – then the fairness assessments could have worked in the direction of “does not deviate unfairly from the actual market conditions” instead of “is unfairly too good or too bad generally.”

The further results depicted in Table 3 give additional impressions regarding what housing conditions people regard as fair. If two children live in the household, a rent of 325 € (or 31 percent) lower than for households without children is considered fair, as well as having 23 m² (or 30 percent) more living space. A good location is regarded as fair if its price is 173 € (or 22 percent) higher than for an average location. Regarding status entitlement, the fair rent for a medical doctor, compared to a nurse, amounts to 252 € more (or 33 percent more), and for a manager, compared to a factory worker, 300 € more (or 42 percent more) per month. If someone is currently unemployed, respondents regard a living space of minus 14 m² compared to a nurse as being fair. Finally, respondents concede 13 m² (15 percent) more living space to those who perform highly professionally. These figures once more illustrate how the need criterion dominates the merit principle: having children goes along with a rent that is reduced by nearly one-third that is regarded as fair (and a 30 percent larger living space), while a high – versus low – job performance only entails a rent reduction of 19 percent (not reported in Table 3) and a 15 percent larger flat.

The final analysis step focuses on the double standards hypothesis with respect to gender and occupation. Remember the baseline effects in this regard: for gender, there was no main effect; regarding occupation, there were main effects in that those with lower-status occupations were judged as entitled to more favorable housing conditions – except for unemployed persons, for whom there was an opposite effect. Table 4 gives an overview of the results from several regression models, interacting the merit, need, and inequality vignette variables with gender and occupation (the full regression tables are in the online appendix). The overall picture is as follows. Firstly, there are double standards with respect to (job) performance, both for gender and for occupation. Secondly, there are only tendencies for gender-specific double standards concerning the need principle (children), the effect is substantially small (see below). Thirdly, the housing inequality indicators (living space, location, monthly costs) do not have differing effects by gender and occupation in terms of justice evaluations; hence, there is no evidence for double standards here. The latter finding is important at it shows that people do not apply different standards for housing costs depending on whether individuals have high- or low-earning occupations.

Going more into detail, Figure 2 depicts predicted values from the regression model interacting gender and (job) performance. The double standard hypothesis assumes diverging effects of job performance by gender and harsher standards for women. The results are in favor of the hypothesis: if women perform poorly in their job (or in a job search if they are unemployed), they are penalized more in terms of what level of residential conditions they merit than if men do so.

Figure 3 shows predicted values resulting from the occupation × performance interaction. Again, there is evidence that people apply different standards concerning job performance, this time depending on occupational status. For physicians and factory workers, performance matters least; for nursing staff and those who are unemployed, it matters most. One could conjecture that concerning low-wage jobs (nurse, factory worker), job performance is evaluated differently in terms of the fairness of housing conditions depending on whether blue-collar jobs or service-oriented jobs are considered. Also, people seem to apply differential standards with regard to how lacking information about job performance is evaluated. For physicians and managers (high-status occupations), no information is taken as equivalent to low performance, while for nursing staff and factory workers (low-status occupations), no information has the same impact as high performance. I want to avoid reading too much into these results (the effects sizes are not always very pronounced and they certainly should be replicated in future studies), but it appears that people have very nuanced notions/perceptions of who merits what in terms of residential conditions when taking occupation and job performance as the basis of judgment. However, there is no clear picture in terms of tougher standards for low-status occupations.

Figure 2:

Conditional Effects Plot: Gender x Performance Interaction

Note: Predicted values from the regression model depicted in Table A10 in the online appendix.

Figure 3:

Conditional Effects Plot: Occupation × Performance Interaction

Note: Predicted values from the regression model depicted in Table A16 in the online appendix.

Figure 4:

Conditional Effects Plot: Gender × Children Interaction

Note: Predicted values from the regression model depicted in Table A12 in the online appendix.

Finally, there are tendencies for gender-specific double standards regarding the need principle. Figure 4 shows that the “penalty” for having no children, compared to having two, is more pronounced for women. Put differently, a man with no children is evaluated as possessing “too good” housing conditions to a lesser extent than a woman with no children. However, the difference is not substantially large and is significant only at the 10 percent level. Not reported separately, and even weaker with respect to statistical and substantial significance, is the tendency that living with a partner entails less surplus in what people see as fair conditions for women than for men. Also not reported separately (see the online appendix) is a statistically significant negative interaction effect between being unemployed and living with a partner. While this indicator for the need principle tends to have no effects for the other occupations, single unemployed persons are rated about 0.4 points more in the direction of “unfairly too good” than those living with a partner.

5 Discussion: Insights, Limitations, and Suggestions for Future Research

The aim of this article has been to investigate what housing conditions people regard as fair for whom, and which distributive justice principles guide their evaluations. The core of the research strategy was a factorial survey (FS) experiment in which justice principles and housing inequality indicators were implemented in vignettes describing example residential constellations, which respondents rated with respect to whether they perceived them as fair. To the best of my knowledge, the article represents the first attempt to investigate fairness of housing evaluations using state-of-the-art (experimental) methods. It contributes both to empirical justice research and to the ongoing debate on (rising) residential inequality. It hence addresses issues that are not only important for scientific debate, but also for the public debate and for policy-makers.

The article has tested three theoretical claims. Firstly, it was argued that we can assume that people do apply the justice principles of merit/equity, need, and status entitlement when making fairness evaluations about different people living in different residential circumstances. Secondly, regarding the question of the relative importance of these justice principles, two competing hypotheses (following Reeskens & van Oorschot 2013) were developed. On the one hand, one can understand housing as a matter of individual, self-inflicted and projectable responsibility which would entail that merit should dominate need. On the other hand, several factors determining housing situations are out of an individual’s control and are of a non-projectable, exogenous character (e. g., the price level in local housing markets, discrimination, limited mobility), which would entail that need (or also equality) should be more important than merit. Thirdly, it was assumed that double standards are at play with respect to gender and occupational status, with tougher standards for women and those with lower-status occupations.

Regarding the first claim, the findings confirm what has been reported for other topics: individuals adopt all justice principles when evaluating unequal residential conditions, though in a nuanced and partly unexpected manner. People performing highly jobwise, or people who are more in need because they have children or live with a partner, are considered to be entitled to more favorable housing conditions. For occupational status, however, I find more or less the exact opposite of what has been established regarding the fairness of earnings: those with high-status occupations are evaluated as having “unfairly too good” housing conditions as compared to those with low-status occupations. The results do not show a gender main effect of the form that men or women are entitled differently favorable housing conditions. The findings further suggest that the need principle is more important than merit, which would favor the second interpretation regarding the above-mentioned competing hypotheses – housing as constituting a non-projectable, exogenous risk. This interpretation is further corroborated by the strong effects of living space and rent level, which dominate once more the merit criterion. It is regarded as highly unfair if people live in very large dwellings or pay excessive rents, which, seen from a broader perspective, could be interpreted as an indicator of the importance of the equality principle: before thinking about who merits what and why, respondents primarily regard these two main indicators of residential inequality as decisive for fairness evaluations. Hence, low housing costs and avoiding exorbitantly large dwellings (by policy measures, for instance) would be important factors for establishing housing regimes that are perceived as fair. Finally, regarding the third theoretic claim, the study finds evidence for double standards dependent on gender and occupational status: the merit criterion has a different impact on what is considered fair depending on these variables and, concerning gender, weak evidence for double standards were also found with respect to the need principle.

A further finding of the study is that people more or less appear to agree in their fairness evaluations: the ICC was zero, and no evidence for random slopes, effects of respondent-level variables, or cross-level-interactions could be found. If this finding holds, it has important implications, as it shows there is a consensus in society in terms of what housing constellations are perceived as fair (however, see some disclaimers on this below). This point, in conjunction with the findings that need dominates merit, and that people regard high rents and too generous housing spaces as highly unfair, leads to the overall interpretation that perspectives on housing lean more to the pro-social and “housing as a basic right” side than to the neoliberal market or “every person for themselves” side.

The study involves several limitations, but it also offers some methodological insights. Certainly, the most important limitation is a design-induced gap between the universe of all factors relevant for evaluating fair housing conditions, on the one side, and what it was feasible to implement in one study (namely, in one FS experiment), on the other. Indicators for equity/merit, need, and status entitlement were limited in number; other possible factors in terms of double standards (and discrimination, which was not addressed at all in this article), such as ethnic origin, had to be excluded. The vignettes did not vary the available income of the example persons, which could have biased the effects of the occupational status dimension. Also limited were the indicators on housing inequality: the design included only three of these, and several others could not be implemented (e. g., homeownership, neighborhood characteristics, building types, condition of the dwelling). All this poses a problem insofar as the methodological literature on FS experiments emphasizes that vignette setups should always contain all relevant information on the issue under investigation (Auspurg & Hinz 2015: 19; Treischl & Wolbring 2022: 157 f.). The design also did not explicitly test for the equality principle. Further, the study only focused on a restricted part of housing markets – namely, the rental market for apartments – and fixed further characteristics in the vignette design. Another point of debate is the local context and the local sample in which this study was conducted, which immediately raises the question of generalizability. Given that this study was conducted in a tight local housing market with extremely high rents, the results are probably not generalizable to a nationwide level. The local context also entails that the sample is rather homogenous – 60 percent of respondents hold a university degree, which could explain the overall consensus found in terms of fairness evaluations. This should be addressed by future work. In turn, restricting fairness of housing studies to geographically limited contexts could also be a methodological advantage: an insight of the study is that, all in all, the FS experiment worked well (with enough variation and low nonresponse rates), and it seems that respondents had no problems dealing with the rating tasks. It has been conjectured above that (presumably) better knowledge about local housing markets facilitates this. One should also bear in mind that the vignette design used here could probably not be directly transferred to a national level, or, more generally, to housing markets whose variance in price levels and living space, as well as other factors, would have made the example constellations in the vignettes unrealistic. Future research should in any case address whether and why fairness of housing evaluations are best studied in geographically limited or broader (national) contexts. Further, the design used in this study focused on individual persons and not on households, but decisions about housing choices and the impacts of residential inequality primarily concern entire households, so concentrating on the household level when studying fairness of housing evaluations could better reflect reality. Another insight is that it seems to be a good idea to use metric vignette variables and to consider their real-world distributions (as has been done for rents and living space in this study). This also allows for estimating illustratively instructive figures on cross-elasticities, as reported above.

In addition to the points already raised, I identify the following research desiderata for future work on fairness of housing research. Generally, the work presented here, and the lack of existing research, calls for more studies on this topic. Future studies should try to replicate the findings of this study and address especially the issues of an ICC of zero (do people generally agree in fairness of housing assessments?), a potentially biased effect of the occupational status indicator (due to missing information on income), a very high square meter price that was regarded as being fair, and ambiguous findings regarding the role of missing information on performance for different occupations. Recall that the findings regarding the occupational status main effects were ambiguous: on the one hand, it is very plausible to suspect that the counter-intuitive negative effects (less favorable housing conditions for higher-status occupations) are a direct result of missing information on income – but in this case, we should have seen an interaction of occupation and housing costs, with the latter being less negative for high-earner occupations, and this was not the case. Future work should also address questions that had to be excluded in this study due to design restrictions. This includes conducting research in less locally limited contexts, namely on a nationwide level. In this regard, it might be reasonable to think of conditional vignette levels that take into account the local housing market characteristics and/or inequality structure (e. g., rural and urban contexts, high- versus low-priced contexts, etc.). Further, this study did not explicitly test for the equality justice principle. A question that should also be addressed by future work is a possible generational conflict in allocating housing resources: how do people perceive the fairness of the fact that, often, elderly people without children live in much better and more spacious housing conditions (and if they are homeowners, at no cost) than young families? Finally, this study is entirely silent about respondent effects on justice evaluations. Future work could investigate well-established theoretical claims in the empirical justice literature in this regard: for instance, the differential norms hypothesis (Auspurg et al. 2017) or the “haves versus have-nots” argument (Reeskens & van Oorschot 2013). This could ultimately also include conducting cross-country comparisons on the role of distributive justice principles and how they are perhaps affected by different country-specific housing regimes or housing policies.