We now move on to the results of our analysis. We examine the relationship between life satisfaction and perceived health status on one hand, and age and geography on the other. We do this for the years 2019 (before the pandemic) and 2020 (during the pandemic). We also conduct sub-analyses where we focus specifically on older adults.
4.1 Life satisfaction—Full sample analyses
The results from the full-sample analysis of life satisfaction in 2019 and 2020 are presented in Table
2. The table displays results for the variables of interest, but we also control for individual characteristics (gender, income, education, marital status, health, housing type, and the month in which the survey was conducted). In addition, independent variables on age, income, and health show significant heteroskedasticity and are included also in a variance equation in the OGLM model. The full table, Table
5 in the Appendix, displays the results for the control variables and variance parameters.
Table 2
OGLM estimates: Life satisfaction (dep. var.)
| (1) | (2) | (3) | (4) | (5) |
Age 16–29: ref. category |
Age 30–59 | 0.927*** | 0.914** | 0.941 | 1.040 | 1.106 |
| (0.026) | (0.038) | (0.047) | (0.144) | (0.158) |
Age 60–69 | 1.114*** | 1.078 | 1.153** | 1.194 | 1.333** |
| (0.037) | (0.052) | (0.069) | (0.171) | (0.191) |
Age 70–85 | 1.401*** | 1.298*** | 1.518*** | 1.373** | 1.919*** |
| (0.054) | (0.078) | (0.113) | (0.205) | (0.308) |
Rural area: ref. category |
Smaller aggl | 0.900*** | 0.934 | 0.867*** | 0.900 | 0.849 |
| (0.030) | (0.042) | (0.044) | (0.146) | (0.149) |
City or larger aggl | 0.889*** | 0.934* | 0.850*** | 1.061 | 1.075 |
| (0.028) | (0.035) | (0.040) | (0.143) | (0.152) |
Metropolitan | 0.865*** | 0.908** | 0.826*** | 1.022 | 0.963 |
| (0.029) | (0.041) | (0.039) | (0.136) | (0.145) |
Year = 2020 | 1.004 | | | | |
| (0.017) | | | | |
Rural area: ref.cat. # Age 16–29: ref.cat |
Smaller aggl. # Age 30–59 | | | | 1.044 | 1.009 |
| | | | (0.185) | (0.192) |
Smaller aggl. # Age 60–69 | | | | 1.006 | 1.055 |
| | | | (0.180) | (0.205) |
Smaller aggl. # Age 70–85 | | | | 1.070 | 1.018 |
| | | | (0.188) | (0.203) |
City or larger aggl. # Age 30–59 | | | | 0.820 | 0.778 |
| | | | (0.116) | (0.124) |
City or larger aggl. # Age 60–69 | | | | 0.885 | 0.808 |
| | | | (0.134) | (0.131) |
City or larger aggl. # Age 70–85 | | | | 0.948 | 0.715** |
| | | | (0.147) | (0.119) |
Metropolitan # Age 30–59 | | | | 0.876 | 0.888 |
| | | | (0.129) | (0.141) |
Metropolitan # Age 60–69 | | | | 0.875 | 0.843 |
| | | | (0.131) | (0.154) |
Metropolitan # Age 70–85 | | | | 0.869 | 0.733* |
| | | | (0.131) | (0.137) |
OGLM variance coefficients | Yes | Yes | Yes | Yes | Yes |
Individual controls | Yes | Yes | Yes | Yes | Yes |
Constants | Yes | Yes | Yes | Yes | Yes |
Pseudo R2 | 0.169 | 0.176 | 0.166 | 0.176 | 0.166 |
N | 18,805 | 8971 | 9834 | 8971 | 9834 |
First and foremost, the results in Table
2, column 1, where we pool the 2019 and 2020 data, convey that subjective life satisfaction is not significantly different in 2020 compared to 2019. The survey question on life satisfaction is broad and represents overall satisfaction with one’s life as a whole.
Secondly, individuals in the age categories of 60–69 years (in 2020) and 70–85 years (in 2019 and 2020) are significantly more likely to report higher levels of life satisfaction compared to the baseline category of 16–29-year-olds across estimations. Individuals aged 30–59 years reported the lowest level of life satisfaction, in accordance with the U-shaped relationship between age and life satisfaction (Columns 1–3). Based on the point estimates, the difference between the age category of 70–85 years and the base category of 16–29 years has increased during the pandemic. This indicates that the gap in life satisfaction between the youngest and oldest age groups has widened in 2020 compared to 2019. However, t-tests show that there is no statistically significant difference in coefficients in 2019 and 2020 (Columns 2–3).
Examining the role of geography, we find that individuals living in rural areas (the baseline category) are more likely to report higher levels of life satisfaction compared to those living in urban areas, both before and during the pandemic (Columns 1–3). Based on the point estimates, the differences were further pronounced during the pandemic, with individuals living in metropolitan areas (Stockholm, Gothenburg, and Malmö) being even less likely to report higher levels of life satisfaction than those residing in rural areas in 2020 compared to 2019. However, t-tests convey that there is no statistically significant difference in the relationship between geography and life satisfaction in 2020 compared to 2019 if we choose a 5% significance level.
The inclusion of geography-age category interaction variables in the estimations in Columns 4 (2019) and 5 (2020) results in the overall relationship between life satisfaction and geography becoming statistically insignificant. Instead, we identify a few significant interaction variables. Based on the results for 2020, we can see that the lower likelihood of reporting higher life satisfaction in urban areas (Columns 1–3) is driven by the older adult population (Column 5). That is, individuals in the age category 70–85 are significantly less likely to report higher life satisfaction levels if they live in metropolitan areas, cities, or larger agglomerations than if they live in rural areas.
During the pandemic year 2020, we see that the older age categories, 70–85 years, who live in urban areas (cities, larger agglomerations, or metropolitan areas) experienced significantly lower levels of life satisfaction compared to their peer group residing in rural regions. This indicates that although older people generally have a higher degree of life satisfaction, during the pandemic in particular, older adults in urban environments experienced lower life satisfaction than older adults in rural areas.
Based on the variance parameters provided by the OGLM (see Appendix Table
5), it is clear that individuals in the reference age category, aged 16–29, are more variable in their life satisfaction scores than older individuals. Furthermore, the higher the household income or subjective health score, the less variable the respondents are in their life satisfaction scores.
We also run ordered logit estimates and the results are robust (see Table
6 in the Appendix). In addition, we add a control for NUTS1 region in the analyses presented in Table
7. There is no significant relationship between life satisfaction and NUTS1. The results on location (rural area, smaller agglomeration, etc.) remain the same. Thus, the density of the location seems to matter more for life satisfaction than where in the country one lives. Table
8 in the Appendix displays results on the relationship between life satisfaction and location, using the Swedish Agency for Growth Analysis’ municipal classification as the determinant of individuals’ type of location. The results are similar to those in Table
2; however, some significance levels vary between the analyses. The lower life satisfaction in urban areas compared to rural areas in 2020 compared to 2019 is more evident when we employ this location variable in the analyses. The gap seems to have widened between the different types of municipalities in 2020 compared to 2019. The interactions between age and location show similar relationships as in Table
2, but they are not statistically significant in Table
8.
Table 3
OGLM estimates: Health status (dep. var.)
| (1) | (2) | (3) | (4) | (5) |
Age 16–29: ref. cat | | | | | |
Age 30–59 | 0.890*** | 0.919* | 0.869*** | 0.828 | 0.849 |
| (0.024) | (0.045) | (0.031) | (0.123) | (0.106) |
Age 60–69 | 0.886*** | 0.927 | 0.855*** | 0.811 | 0.782** |
| (0.030) | (0.048) | (0.037) | (0.139) | (0.098) |
Age 70–85 | 0.696*** | 0.696*** | 0.695*** | 0.704** | 0.599*** |
| (0.028) | (0.043) | (0.037) | (0.125) | (0.085) |
Rural area: ref. cat |
Smaller aggl | 0.964 | 0.966 | 0.961 | 0.793 | 0.883 |
| (0.030) | (0.052) | (0.041) | (0.136) | (0.122) |
City or larger aggl | 1.048 | 1.018 | 1.068* | 0.955 | 0.978 |
| (0.032) | (0.050) | (0.041) | (0.141) | (0.112) |
Metropolitan | 1.128*** | 1.103* | 1.144*** | 1.016 | 1.092 |
| (0.036) | (0.064) | (0.046) | (0.165) | (0.138) |
Year = 2020 | 0.961** | | | | |
| (0.018) | | | | |
Rural area: ref. cat. # Age 16–29: ref. cat | | | | | |
Smaller aggl. # Age 30–59 | | | | 1.210 | 1.012 |
| | | | (0.231) | (0.156) |
Smaller aggl. # Age 60–69 | | | | 1.374 | 1.201 |
| | | | (0.292) | (0.188) |
Smaller aggl. # Age 70–85 | | | | 1.199 | 1.237 |
| | | | (0.259) | (0.208) |
City or larger aggl. # Age 30–59 | | | | 1.104 | 1.060 |
| | | | (0.169) | (0.138) |
City or larger aggl. # Age 60–69 | | | | 1.118 | 1.093 |
| | | | (0.202) | (0.150) |
City or larger aggl. # Age 70–85 | | | | 0.940 | 1.203 |
| | | | (0.179) | (0.169) |
Metropolitan # Age 30–59 | | | | 1.134 | 0.980 |
| | | | (0.177) | (0.136) |
Metropolitan # Age 60–69 | | | | 1.148 | 1.077 |
| | | | (0.196) | (0.150) |
Metropolitan # Age 70–85 | | | | 0.918 | 1.209 |
| | | | (0.171) | (0.213) |
OGLM variance parameters | Yes | Yes | Yes | Yes | Yes |
Individual controls | Yes | Yes | Yes | Yes | Yes |
Constants | Yes | Yes | Yes | Yes | Yes |
Pseudo R2 | 0.076 | 0.079 | 0.073 | 0.079 | 0.073 |
N | 18,805 | 8971 | 9834 | 8971 | 9834 |
Table 4
OGLM estimates: Life satisfaction and health status (dep. var.). Restricted sample 70 years old and above
| (1) | (2) | (3) | (4) |
Age 70–75: ref. category |
Age 76–80 | 1.031 | 1.132* | 0.800*** | 0.797** |
| (0.080) | (0.079) | (0.065) | (0.084) |
Age 81–85 | 1.178 | 1.367*** | 0.593*** | 0.654*** |
| (0.131) | (0.135) | (0.079) | (0.090) |
Rural area: ref. category |
Smaller agglomeration | 0.925 | 0.859 | 0.993 | 1.095 |
| (0.107) | (0.100) | (0.126) | (0.121) |
City or larger agglomeration | 0.954 | 0.722*** | 0.929 | 1.241* |
| (0.097) | (0.084) | (0.113) | (0.141) |
Metropolitan | 0.768** | 0.651*** | 0.970 | 1.422** |
| (0.101) | (0.077) | (0.118) | (0.216) |
OGLM variance parameters | Yes | Yes | Yes | Yes |
Individual controls | Yes | Yes | Yes | Yes |
Constants | Yes | Yes | Yes | Yes |
Pseudo R2 | 0.186 | 0.157 | 0.082 | 0.064 |
N | 1936 | 2099 | 1936 | 2099 |
In summary, older individuals and those living in rural areas reported higher levels of life satisfaction before and during the pandemic, with this pattern being slightly reinforced during the pandemic year. In 2020, older adults in urban areas had significantly lower life satisfaction compared to older adults in rural areas. Finally, the type of area one lives in seems to matter more for life satisfaction than where in the country one lives.
Our finding that overall life satisfaction has not been severely affected by the pandemic in the short run is in consensus with previous literature, which emphasizes that subjective life satisfaction is not a quick or temporary judgment based on momentary influences and feelings but a relatively stable phenomenon (e.g., Pavot et al.
1991). Presumably, individuals base their life satisfaction on perceptions of their past and future and believe that the pandemic will not have long-term consequences on their life situations.
Taken together, our results suggest that younger individuals may have experienced a different kind of stress and uncertainty than older individuals, albeit to a somewhat lesser extent from public restrictions. For instance, high school and university students received their education entirely remotely. Additionally, other age groups may have experienced greater stress due to potential negative effects on their employment opportunities, while the older generation’s pensions remained unchanged.
One reasonable explanation for the fact that older adults in urban environments experienced lower life satisfaction than those in rural areas may be that the urban lifestyle was more significantly affected by the restrictions. For example, older adults in urban areas presumably had a harder time going outdoors while following the recommendations and may therefore have stayed indoors and more isolated than their rural counterparts. Higher population density in urban areas may also have caused more fear of being infected. Furthermore, it is more difficult to maintain distance on public transport or to get around by car. Additional factors could include variations in the sense of community or living conditions (size of accommodation/garden).
4.2 Health—full sample analyses
Turning to our results on the relationship between subjective health, age, and geography before and during the pandemic, we observe a somewhat different pattern. The results from the full-sample analysis of health in 2019 and 2020 are presented in Table
3. As in Tables
2,
3 displays results for the variables of interest, but we also control for individual characteristics (gender, income, education, marital status, life satisfaction, housing type, and the month in which the survey was conducted) and heteroskedasticity for the variables significant in Brant tests: age, geography, income, life satisfaction, education, gender, and housing type. The full table displays the results for the control variables and variance parameters (see Appendix Table
9).
In contrast to the results on life satisfaction, the results in Table
3, column 1, where we pool the 2019 and 2020 data, convey that subjective health is significantly lower in 2020 compared to 2019. The pandemic appears to have had a significant effect on individuals’ perceptions of their health status; however, we cannot distinguish whether these perceptions correspond to actual declines in “objective” health conditions.
Furthermore, all older age categories are more likely to report worse health status compared to the younger age group (base category 16–29 years), which is expected. However, the age groups 30–59 and 60–69 report similarly lower levels of health scores relative to the reference category, while the age group 70–85 reports significantly lower levels.
In columns 1 and 2, we investigate the relationships between health, age, and geography separately for the years 2019 (column 2) and 2020 (column 3). In relative terms, the health gap between the oldest and the youngest age categories remains similar in the two years. On the contrary, based on the point estimates, the age gap between middle-aged individuals and the youngest individuals appears to have increased slightly in 2020 compared to 2019 (the corresponding 2019 and 2020 coefficients are not statistically different from each other in t-tests). Additionally, we find that individuals residing in metropolitan areas (2019 and 2020) and cities or larger agglomerations (2020) are more likely to report higher health scores compared to those living in rural areas.
In order to gain further insight into the relationship between age and location in relation to health, we include interaction variables between age and place of residence in columns 4 (2019) and 5 (2020), while controlling for the same individual characteristics. The results of our analysis reveal that the relationship between age and health remains consistent with our earlier findings, with older individuals reporting lower perceived health than the reference group (16–29 years), especially during the pandemic year of 2020. When we include these interaction variables, the relationship between geography and health becomes insignificant, and the negative relationship between the two oldest age categories and health becomes larger in magnitude in 2020. While the interaction variables are statistically insignificant, the point estimates in Table
3, column 5, suggest that the geographical variation in subjective health score is driven by older adults in urban areas perceiving their health to be better than their peers in rural areas.
Based on the variance parameters provided by the OGLM, we find that individuals aged 70–85 are more variable in their health scores than younger individuals (reference category, 16–29). Furthermore, individuals in smaller agglomerations, cities, or larger agglomerations are less variable in their health scores compared to individuals in rural areas (full results available in Appendix Table
9). We also run ordered logit estimates and the results are robust (see Table
10 in the Appendix). In Table
11 in the Appendix, we add a control for the NUTS1 region in the analyses. There is a slight positive relationship between living in the South (rather than the North) and health status in 2019 but not in 2020. No other NUTS1 coefficient is significant, and the results for the other variables remain the same when this variable is controlled for. Thus, the rurality/density of the location seems to matter more, both for health and life satisfaction, than where in the country one lives. In Table
12 in the Appendix, the Swedish Agency for Growth Analysis’ municipal classification is used as a determinant of individuals’ type of location. The results are similar to those in Table
3; however, the relatively higher health score in urban areas compared to rural areas in 2020 compared to 2019 is more evident when we employ this location variable in the analyses. The gap seems to have widened between the different types of municipalities in 2020 compared to 2019. The similar results in Tables
2 and
8, and Tables
3 and
12, employing different measures of type of location (self-reported location and municipal classification), indicate that the results are robust.
In summary, older individuals and those living in rural areas reported lower perceived health status, which was further reinforced during the pandemic year.
4.3 Life satisfaction and Health – sub-sample analyses
To investigate more in depth the relationship between geography/location and well-being for older adults, we run a subsample analysis including only individuals aged 70–85 years. This sub-analysis allows us to examine if the overall relationships between geography and well-being differ for this age group. The results are displayed in Table
4 (full table available in Appendix Table
13). We also control for age in this subsample, using the following age categories: 70–75, 76–80, and 81–85, with 70–75 forming the reference group.
Starting with the results on life satisfaction in the years 2019 (column 1) and 2020 (column 2), we can see that there is significant variation in life satisfaction between individuals aged 70–75 and older individuals in 2020. That is, individuals 70–75 are significantly less likely to report higher life satisfaction in comparison to individuals 76–85 during the pandemic but not before the pandemic. In consensus with the findings in Table
2 (i.e., the geographical variation in life satisfaction is driven by the variation in life satisfaction of older adults across space), the results in Table
4 display that older adults living in metropolitan areas (2019 and 2020) and cities or larger agglomerations (2020) are significantly less likely to report higher life satisfaction. In other words, in 2019, older individuals in the metropolitan areas (Stockholm, Gothenburg, and Malmö) felt that they had a lower degree of life satisfaction than the reference group who lived in the countryside. During the pandemic year, this also applied to those who lived in other urban environments (cities and larger agglomerations). That is, the older age group of 70–85 years living in urban environments experienced a lower degree of life satisfaction during the pandemic in comparison to 2019 (the difference in coefficients between 2019 and 2020 is statistically significant in t-tests, p-value = 0.07).
Table
4, columns 3 (year 2019) and 4 (2020), display the results on the relationship between subjective health, age, and geography. The results show a significant negative relationship between age and health, as expected, and this relationship is similar in 2019 and 2020. Regarding geography, we do not find any significant results for the year 2019, i.e., the year before the pandemic, for these older individuals. However, in the pandemic year 2020, we observe a significantly higher likelihood of higher perceived health in urban settings compared to rural areas.
It is possible that the restrictions had a more pronounced effect on the age group 70–75 than on those who were older, as individuals in the younger group are more likely to engage in a wide range of cultural and social activities compared to their older counterparts. From a geographical perspective, it could be that people in cities had closer contact with healthcare during the pandemic, as accessibility to healthcare is significantly better in cities than in rural areas. Although it is unlikely that the actual health status of urban older adults improved compared to rural older adults within a year, it is clear that perceived health status was higher in urban areas than in rural areas in 2020 compared to 2019. This may have been a result of the frequent emphasis on the connection between being older and being in a higher-risk group in 2020, which led older individuals in rural areas to perceive themselves as even less healthy than those in cities at a time when health status was in focus. We also ran the analysis using an ordered logit regression, and the results are robust (see Appendix Table
14).