Introduction

The difference between urban and rural areas, in terms of standard of living and subjective well-being, is still being debated. If, on one hand, cities offer higher job opportunities, leisure activities and cultural events, on the other hand, they are characterised by higher cost of living, higher levels of pollution and greater wealth inequality. Thus, existing evidence is inconsistent and heterogeneous across time and countries. In early 2020, the COVID-19 pandemic caused a global health crisis, infecting more than 210 million people and claiming more than 4 million lives in a couple of years (Worldometer, 2021). France was one of the first countries implementing measures to keep physical distance between individuals. Since public spaces were closed for a long time, the COVID pandemic, requiring everyone to be locked down at home, might have exacerbated the impact of living spaces on individuals’ quality of life by widening urban–rural differences in subjective well-being.

By using a probability-based panel study, consisting of 1,404 individuals, we explored changes in subjective well-being over time, from the pre-pandemic period (2019) to about one year into the pandemic (April 2021). In addition, we investigated between-individuals differences based on rural–urban differential factors (i.e. compositional factors) and within-individuals differences based on events that have been experienced during the pandemic period (i.e. contextual factors). Quantitative findings suggest a short-term improvement of subjective well-being compared with the pre-lockdown period. Indeed, net of individual socio-economic and demographic characteristics, subjective well-being had slightly increased in the first phase of the lockdown, showing a peak in the summertime (end of May–early June), but a strong decline in the autumn, returning to initial values of the pre-lockdown phase, to increase again in the second lockdown phase (April 2021). By looking at changes of subjective well-being across different degrees of urbanisation, we have first distinguished rural areas or small towns, suburbs, small cities and large cities. Findings have generally revealed higher levels of subjective well-being in small areas, while individuals living in large cities and, especially, in suburbs were among those who reported the lowest level of subjective well-being.

To identify potential mechanisms to explain this gap, we have first looked at between-individuals differences of subjective well-being, due to a set of compositional factors. Findings suggest that material conditions, social capital and physical environment only partially account for changes of subjective well-being over time and its difference across residential areas. Then we looked at within-individuals differences in subjective well-being over time according to specific factors associated with the COVID-19 lockdown. Findings suggest that, controlling for individual fixed effects, having had COVID-19 in the first period of lockdown (early April 2020) decreased subjective well-being among respondents, especially if living in cities, while having helped neighbours increased subjective well-being.

Subjective well-being in rural and urban areas

The degree of urbanisation has been widely associated with economic growth and higher living standards. Therefore, many studies have also argued that living in cities would be associated with higher levels of happiness (Glaeser, 2011; Burger et al., 2020). However, with the generalised increase of wealth and the development of technology since World War II, the urban–rural differentials in happiness and well-being might have been reduced or even eliminated. Indeed, the individual’s average level of well-being in large cities has declined mainly because the pros due to urbanisation (economic development, job opportunities, etc.) are associated with higher costs of living, higher pollution, lower level of social capital and inequality. Thus, it is unclear why, although living in urban areas is associated with lower levels of life satisfaction (Rodríguez-Pose & Maslauskaite, 2012), most people are migrating towards cities. This phenomenon has been called ‘the urban happiness paradox’ (Sørensen, 2021). The reasons behind higher levels of well-being in rural contexts have been extensively investigated. For example, some research has suggested that small town are characterised by informal social contacts and a homogeneous population, which facilitate stronger social networks and good psychological health (De Vos et al., 2016). According to Hoogerbrugge and Burger (2020), other reasons can be drawn from a range of socio-economic, contextual and environmental factors since cities have higher levels of air pollution, noise of cars and public transport and a lack of green space. Finally, compared to smaller cities, large urban areas often have higher levels of poverty and inequality (Graham & Felton, 2006).

Having said this, for most parts of the world, there is no evidence that either rural or urban areas are associated with significant variations in happiness (Berry & Okulicz-Kozaryn, 2009). The bulk of researchers agree on arguing that personal characteristics and level of development are the key driving forces of subjective well-being. However, the COVID-19 pandemic, requiring everyone to be locked down at home for an unusually long time, might have widened urban–rural differences in subjective well-being, revealing the strength and the weakness of both compositional and contextual factors. For example, being locked down in small houses and without any green spaces (i.e. in large cities) might have unequally changed subjective well-being across individuals. With public spaces closed, Paris might be a worse place to be locked down than the average French town or rural residence (see Recchi et al., 2020). Finally, changes in individuals’ condition (i.e. having been infected with COVID19) and behaviours (i.e. work from home; having helped others with basic needs) might be a different experience in urban vs rural regions leading to different effects on subjective well-being.

Urban and rural differences in subjective well-being under the COVID-19 crisis in France: compositional and contextual factors

The outbreak of the SARS-CoV-2 virus and the associated coronavirus disease 2019 (COVID-19) originated in China in December 2019 and quickly spread across the globe causing more than 4 million deaths up to August 2021. According to Evandrou et al. (2021), the lockdown brought disruption to daily life for the whole population and measures adopted to contain the virus are likely to leave their mark by producing detrimental effects on financial, relational, physical and psychological domains. However, in some cases, the perception of both health and well-being during the COVID-19 epidemic has improved in comparison to previous years (Recchi et al., 2020). This phenomenon has been called ‘the eye of the hurricane’ paradox and it argues that most individuals, who have not been infected by the virus, felt happier and in better health then they would normally do (Recchi et al., 2020). This is what has been previously found after the Great East Japan Earthquake, suggesting that the general perception of well-being in life increased following the disaster. According to Uchida et al. (2014) this can be explained not only by the fact that normal factors lose their power in influencing well-being after a disaster, but also because people change their expectations of their life after a tragic event.

Indeed, feeling that one’s own condition is more favourable compared with that of others may lead individuals to report higher levels of subjective well-being (Schwarz & Strack, 1999). However, although the majority of people have declared feeling healthier and having a better feeling than before the lockdown, this trend was not equal across social classes (Recchi et al., 2020). Indeed, empirical findings showed that the pandemic has exacerbated health disparities as individuals reporting lower levels of well-being were consistently those belonging to the working class and the most financially vulnerable, people living alone and in smaller homes, those who were not born in France, and women (Recchi et al., 2020; Schradie et al., 2020). Preliminary findings on the same data set (Recchi et al., 2020) have also shown that residents in Paris experienced a significant decrease in their subjective well-being score compared to the rest of the country.

Data and method

Like Recchi et al. (2020), this study used nationally representative panel data of French residents. Specifically, we have drawn seven survey waves from ELIPSS, which is a probability-based panel launched in 2012 and managed by the CDSP (Center for Socio-Political Data of Sciences Po). The sample consists of 1,404 French residents that have been initially drawn from census data. The average response rate was of about 85 per cent (Recchi et al., 2020). The first survey was administered two weeks after the start of the lockdown (April 1–8, 2020). Subsequent waves were carried out at two-week intervals: the second (April 15–22), the third (April 29–May 6), the fourth (May 23–20) and the fifth (May 27–June 4). The fourth and fifth waves came after the end of the first lockdown (on May 11) and during the period of economic reopening. The sixth wave (22–29 October) was administered during the deconfinement period just before the second lockdown, when the seventh wave was administered (19–26 November). Finally, a last wave was collected one year after the beginning of the pandemic (22–29 April 2021). The baseline of the current study is the ELIPSS annual survey carried out in 2019.

Post-stratification weights based on sex, age, education and region of residence have been computed to account for design effects and possible bias due to attrition and the acceptance rate in the enrolment phase (about 25 per cent). The collected information includes individual physical and mental health status, subjective well-being, working conditions, daily living activities and specific questions about changes due to the pandemic lockdown.

With regard to subjective well-being, we developed an index allowing us to capture subjective well-being in a holistic way (Recchi et al., 2020). More specifically, we combined respondents’ responses to seven different questions regarding how often they had felt ‘nervous’, ‘low’, ‘relaxed’, ‘sad’, ‘happy’, ‘in good health’ and ‘lonely’ over the previous two-week period on a five-point scale from never to always. Negative feelings are inverted to enable the construction of an additive index, which was then normalised between 0 (lowest) and 1 (highest).

The degree of urbanisation has been defined according to the French census’s urban unit definition (i.e. Tranche d’unité urbaine 2014) which distinguishes rural and urban areas according to the number of inhabitants. More specifically, the variable has been recoded according to Berry and Okulicz-Kozaryn (2011)’s thresholds, taking value 1 for rural areas or small towns (<10,000 inhabitants), 2 for suburbs (between 10,000 and 50,000 inhabitants), 3 for small central cities (>50,000 and fewer than 200,000 inhabitants) and 4 for large central cities (>200,000 inhabitants).

Unadjusted descriptive statistics show the distribution of subjective well-being index over time according to the respondents’ degree of urbanisation (Figure 20.1). Overall, we observe an increase of subjective well-being over the first three months of lockdown (April–June 2020), especially for respondents living in rural areas and in small towns compared with the pre-lockdown period. The subjective well-being reported by respondents living in large central cities was almost stable over time, becoming lower compared with those reported in rural areas at the beginning of the lockdown but getting back to initial values at the end of the first lockdown (November 2020). Descriptive statistics have been confirmed by adjusted multivariate regression analyses, suggesting an increase of subjective well-being for those living in small cities (Figure 20.2).

Figure 20.1 Subjective well-being along the rural–urban continuum. (a) Weighted mean of subjective well-being over time and degree of urbanisation. (b) Distribution of respondents living in rural or small towns, suburbs, small central cities or large central cities. (c) Weighted mean of subjective well-being over time and by residence area. (d) Distribution of respondents by residence area. N=1,278; N=8,863 (persons-year).

Figure 20.2 Linear prediction of subjective well-being by (a) degree of urbanisation and (b) time. We regress our subjective well-being index on an interaction with a dummy variable equal to one if the individual resides in rural, towns or urban areas during the lockdowns. The figure plots the coefficients on these interactions. We fix the 2019 level (Enquête Annuelle 2019) as the reference category. We also include the individual controls. N=1,278; N=8,863 (persons-year).

Compositional factors

Evidence about differences in well-being across degrees of urbanisation might be explained by clustering the risk factors of individuals living in the same areas (compositional effect, see Stafford & McCarthy, 2005) to look at the specific characteristics of these locations. Thus, economic, social and environmental characteristics have been considered in order to dig deeper into the drivers of urban–rural differences in subjective well-being, before and during the lockdown. Namely, we used per-person house size and reliability of Internet connection as proxies for respondents’ material living condition. Moreover, social relations may produce externalities potentially influencing individual well-being. Thus, although during the lockdown physical interactions significantly declined (Arpino et al., 2021), we have operationalised social capital through the individual propensity of trusting others. Residential physical environment also matters to identify whether urban–rural differences are driven by compositional factors. Thus, we included in the analysis a variable measuring whether respondents report a lack of green spaces in their neighbourhood and whether their residence space lacks public transport, making them isolated.

A stepwise OLS regression model included these variables as controls (model 1). Then, we interacted all these variables with time (model 2) to check whether these associations changed during the lockdown phases. Finally, controlling for timing, we interacted them with degrees of urbanisation/residential area (model 3). Since the dependent variable is standardised, marginal effects of each independent variable have been calculated by multiplying the coefficient by the standard deviation of the outcome variable (β * σy). By including compositional factors, the difference between subjective well-being reported by residential areas has been slightly modified, suggesting that compositional factors only partially contribute to explain the rural–urban gap.

Living in a house with less than 25m² per person was associated – on average – with a decrease of about 0.003 points of subjective well-being (p<0.01). However, this effect was stronger in both rural areas and large cities (Figure 20.3, panel a) as well as in November 2021 – during the second lockdown – (Figure 20.3, panel b). Having a bad Internet connection was not significantly associated with subjective well-being. However, interaction terms show a negative effect among individuals living in large cities (Figure 20.3, panel c).

Trusting others was positively associated with well-being, although its effect on subjective well-being was smaller in small cities (Figure 20.4, panel a) and decreased over time during the lockdown (Figure 20.4, panel b). Lack of green spaces in the neighbourhood of residence was negatively associated with individual subjective well-being (p-value <0.001) and the effect is even stronger if the respondent lives in suburbs (Figure 20.5, panels a,b). The availability of transport facilities, and therefore the lack of isolation, was particularly positively associated with subjective well-being, especially among those living in cities compared with rural areas (Figure 20.5, panel c; this effect was particularly relevant during the deconfinement phase (October 2021) and during the second lockdown when people returned to the workplace and schools were opened (Figure 20.5, panel d).

Figure 20.3 Linear prediction of subjective well-being by compositional factors (material conditions) and time. We regress our subjective well-being index on an interaction with variables related to compositional factors during the lockdowns. The figure plots the coefficients on these interactions. We fix the 2019 level (Enquête Annuelle 2019) as the reference category. We also include the individual controls. N=1,278; N=8,863 (persons-year).

Figure 20.4 Linear prediction of subjective well-being by compositional factors and degree of urbanisation/residence region. We regress our subjective well-being index on an interaction with variables related to compositional factors for degree of urbanisation/residence region. The figure plots the coefficients on these interactions. We also include the individual controls. N=1,278; N=8,863 (persons-year).

Figure 20.5 Subjective well-being by lockdown phase. We regress our subjective well-being index on survey wave by controlling for individual fixed effects. We fix the 2019 level (Enquête Annuelle 2019) as the reference category. N=1,278; N=8,863 (persons-year).

Contextual factors

Since most of the variation in subjective well-being is often attributable to individual characteristics (Bellas & Tranmer, 2012), we adopted an individual fixed effects approach which, by controlling for unobserved individual time-invariant features, ensures that our estimates are not suffering from selection bias as variables that vary between individuals but not within persons are excluded from the model.

Figure 20.6 shows the linear prediction of subjective well-being by points in time, controlling for individual fixed effects. As already shown by Recchi et al. (2020), individuals’ subjective well-being significantly increased in 2020 compared with that reported in 2019. However, it came back to pre-pandemic values over the course of the lockdown. More specifically, we observe that subjective well-being scores have risen since the quarantine started (1–8 April 2020). The rise is slight but constant over the first two months of lockdown, with a peak at the beginning of June. Then it decreased, and by November 2020 it was almost back to the pre-lockdown values, but it increased again at the beginning of the second lockdown (April 2021).

Figure 20.6 Subjective well-being by lockdown phase and COVID-19 infection. We regress our subjective well-being index on survey wave, having had COVID-19, having kin who had COVID-19 and by controlling for individual fixed effects. We fix the 2019 level (Enquête Annuelle 2019) as the reference category. N=1,278; N=8,863 (persons-year).

Changes of individuals’ subjective well-being over the course of the pandemic might be significantly different according to whether respondents have been infected with COVID-19, changed working conditions, or helped neighbours (i.e. contextual factors). Estimates have been performed, stratifying by degree of urbanisation. Namely, we notice that the increase of subjective well-being over the lockdown period was mostly reported if respondents have not been directly affected by the virus, especially if they live in suburbs and cities (Figure 20.7). Teleworking partially moderated this relationship by increasing well-being during the first phase of the lockdown, but only among respondents living in suburbs (Figure 20.8). Finally, respondents living in large cities who have helped neighbours during the first period of lockdown reported significantly higher scores of subjective well-being (Figure 20.9).

Figure 20.7 Subjective well-being by lockdown phase over COVID infection by degree of urbanisation. N=1,278; N=8,863 (persons-year).

Figure 20.8 Subjective well-being by lockdown phase over changes in working condition by degree of urbanisation. N=1,278; N=8,863 (persons-year).

Figure 20.9 Subjective well-being by lockdown phase over help offered to neighbours by degree of urbanisation. N=1,278; N=8,863 (persons-year).

Conclusions

This study aims to explore whether the COVID-19 pandemic, by requiring everyone to be locked down at home for a long time, has widened urban–rural differences in subjective well-being. Results have highlighted meaningful differences of subjective well-being across residence regions and population density by suggesting that being locked down in small cities was a less stressful experience than in other residential contexts such as large cities and suburbs.

Our empirical findings show that pre-existing material conditions, social capital and physical environment only partially contributed to explain the differences in subjective well-being across residential areas. Per-person house dimensions, the availability of green spaces, having a good Internet connection, lack of isolation, and trust in others were specifically relevant for well-being of those living in large cities but do not fully explain why small cities were better off. Similarly, controlling for individual fixed effects, contextual changes only partially contributed to explain changes in subjective well-being across different residential areas. These findings are not only consistent with existing literature (i.e. Okulicz-Kozaryn, 2017), arguing that although small cities show greater levels of well-being regardless of specific compositional and contextual factors, they also indicate that small cities have reacted better to the pandemic than other locations.

However, our findings also show that respondents living in suburbs and in large cities were those reporting the lowest levels of well-being. Indeed, although the city offers more attractive job opportunities and more social and cultural outlets for young people (i.e. restaurants, clubs, museums, theatres and so on), according to Wirth’s theory (1938), it is also characterised by anomie and alienation. During the pandemic, the availability of spaces (in- and outdoor) and the possibility to be (albeit virtually) connected to others assumed a greater importance to preserve mental health (Corley et al., 2021; Arpino et al., 2021).

This study contributes to the existing literature in many ways. First, it used original panel data able to follow individuals over one full year of the pandemic (April 2020–April 2021). Second, it used a holistic index to assess subjective well-being, able to capture in a more comprehensive way individuals’ feelings. Finally, by using a four-point ordinal scale to measure the urban–rural gradient, we provided original evidence on both compositional and contextual factors influencing the difference of quality of life in urban and rural areas over the pa ndemic.