The consequences of COVID-19 on older adults: evidence from the SHARE Corona Survey

1.   Introduction

The spread of COVID-19 has resulted in a significant slowdown in economic activities on a global scale. According to World Bank (2021) estimates, global GDP decreased by 3.5% in 2020, compared to the previous year. Similarly, OECD (2021) data indicate a 3.4% overall decrease in 2020 with respect to 2019, which reaches 4.7% for advanced economies, while the latest data from IMF (2021) show an overall decrease in global GDP of 3.2% and of 4.6% for advanced economies, with a positive rebound in the latter in the third and fourth quarter of 2020, as a result of the end of the lockdowns in May and June 2020. This was largely due to unprecedented responses by the governments of these countries on fiscal, monetary, and regulatory aspects, which facilitated the maintenance of household disposable income, protected companies' cash flow, and supported the availability of credit (Danielli et al., 2021).

This paper aims at examining the economic consequences of COVID-19 on the European population, particularly focusing on the older age groups (over 50 years of age), making use of the SHARE Corona Survey (Börsch-Supan et al., 2013; Börsch-Supan, 2021). Awareness of the different impact of COVID-19 on the various socio-demographic groups and, in particular, of the double burden (health and economic) borne by high-risk groups makes the assessment of the economic impact of the pandemic on the older population of primary importance (Antipova and Momeni, 2021; Gallego et al., 2021; Gietel-Basten et al., 2022; Pant and Subedi, 2020), especially in a context in which Europe is faced with the challenge of an increasingly ageing population (Cristea et al., 2020).

Our research focuses on three areas, linked by the common concept of economic insecurity (see, e.g., Giambona et al., 2022; Panarello, 2021; Rohde and Tang, 2018). First, we examine the reductions in the well-being of the population under study at the end of the first COVID-19 wave, through respondents' statements on the possibility of satisfying their needs through their current income, trying to identify the contextual factors that make it particularly difficult to achieve this goal. Then, we take into consideration the eventual job loss recorded at the end of the first COVID-19 wave, trying to identify the most significant elements that determine it. Finally, we focus on the share of the population who received financial aid from the State, employer, relatives and/or friends during the first COVID-19 wave, trying to highlight the main economic and social circumstances that led to its provision.

The economic consequences of the pandemic will be great and uncertain, with different effects on the labour market, production chains, financial markets, and GDP levels (Brodeur et al., 2021). The negative effects may differ according to the stringency of the social distancing measures, the duration of their implementation, and the degree of citizens' compliance (Panarello and Tassinari, 2022). Furthermore, the pandemic and its related interventions may have led to a greater spread of mental health disorders (Busetta et al., 2021; 2022; Vaculíková and Hanková, 2021), increased economic inequality, although mitigated by governmental support schemes (Aspachs et al., 2021), and particularly harsh effects on some socio-economic groups, such as older adults (Christensen, 2021).

The pandemic has caused disruptions to the supply system at the local, regional, and global level (Gunessee and Subramanian, 2020; Ivanov and Dolgui, 2020; Majumdar et al., 2020; Paul and Chowdhury, 2020); the repercussions on local and sectoral demand have caused global demand to retreat. The social distancing measures necessary to contain the spread of the SARS-CoV-2 virus have cut demand, especially in the tourism, travel, and hospitality-related services sectors (Kaushal and Srivastava, 2021; Tsionas, 2020). Consumer and business confidence has dropped (Brodeur et al., 2021; Teresiene et al., 2021). Commodity prices have fallen, as a result of both lower global demand and the decision taken in March 2020 by oil-producing countries to increase production (Barichello, 2020; Ezeaku et al., 2021). Destruction of supply and weakening demand have negatively impacted employment and growth, reduced government revenues and imposed further deterioration on public finances, with high debt and associated vulnerabilities which restrict the ability to exercise fiscal support for the economy in many countries (Brodeur et al., 2021).

Indeed, the pandemic has had devastating health and economic effects. According to Eurostat (2021a), the EU countries saw a 6.0% decrease in GDP per capita in 2020 and a 6.9% decrease in consumption (at current prices) by families and private social institutions. The decrease in GDP was particularly intense in the second quarter of 2020 (–8.2%) with a stabilisation in the rest of 2020 and a slow recovery in the first two quarters of 2021. The decrease in employment income during 2020 was, on the contrary, quite modest (–0.7%). The unemployment rate in the Euro area increased from 7.5% to 7.8% during 2020 and the employment rate, compared to 2019, decreased from 73.1% to 72.4%. At the same time, deaths rapidly increased across Europe; already at the beginning of the pandemic, in some parts of Europe, the number of deaths was excessively high compared to the average mortality of the previous periods. Hence the idea of measuring the impact of COVID-19 by looking at excess mortality, i.e., the increase in the total number of deaths for any cause compared to the same period of the previous years. In total, there were over 580,000 excess deaths in the EU between March and December 2020, compared to those that occurred in the period 2016–2019. The pandemic and its economic consequences have also caused a major increase in fiscal deficits and debt-to-GDP ratio in all countries. In particular, in the European Union, the overall deficit in 2020 stood at 6.9% of GDP, compared to 0.5% in 2019.

The main objectives of governments' actions have been to save lives, contain the spread of the virus, cure those who got infected, and protect citizens and businesses from the economic crisis resulting from the pandemic (Brodeur et al., 2021; Panarello and Tassinari, 2022), through reinforced unemployment benefits, wage subsidies, income support, and social assistance, while limiting business closures and bankruptcies in the areas and sectors more at risk. Such actions prevented the health crisis from generating long-term weakness in demand and from reducing the population's well-being. Public policies have had a relevant effect in tempering the consequences of the crisis deriving from COVID-19 on the population's standard of living (Padhan and Prabheesh, 2021). Data from Eurostat (2021b) indicate that the median income from work fell by 7% during 2020, while the median disposable income remained almost unchanged.

There is a strong correlation between health and economic conditions (Mackenbach, 2019). In recent decades, the population's average health conditions have improved in many countries worldwide, leading to decreased mortality and increased life expectancy; the most recent developments in biomedical knowledge continually seem to promise unstoppable progress in this area. However, the growth of socio-economic inequalities, largely determined by the dominant economic and productive models in nowadays societies, was accompanied by a similar increase in health inequalities (Abeliansky and Strulik, 2019). On average, health improves and mortality decreases, but this mostly occurs in the strongest social groups. On the contrary, the former worsens and the latter increases, or, at least, one does not improve and the other does not decrease, in the weakest groups from an economic, social, and cultural point of view: thus, health inequalities increase or, at most, remain stable. Health conditions, in turn, affect economic conditions, determining additional differential needs and decreasing the ability to work and, consequently, income (Smith, 1999).

We hypothesise, as discussed by Gilligan et al. (2020), that a relevant element in determining households' ability to cope with adverse economic situations is given by family and friends networks (H1), even though there might be an inverse relationship, for which households in difficult economic conditions tend to rarefy their social contacts.

Moreover, we hypothesise (H2) a direct relationship between frequency of social contacts and stated health level (Assari, 2017; Minkler et al., 1983).

Our third hypothesis (H3) is that respondents who lost their employment due to the consequences of the pandemic are less likely to be able to make ends meet, compared to those who were not employed even before the outbreak.

However, we also expect such needy respondents to obtain financial support (H4): indeed, those who are poorly able to make their ends meet should get adequate assistance.

Finally, we hypothesise that respondents from countries exhibiting a higher GDP growth, or a lower decrease in GDP, are more likely to be able to meet their household's expenses (H5).

The remainder of the present manuscript is structured as follows. Section 2 portrays the data, providing some descriptive analyses and a description of the estimated models. Then, Section 3 presents the results from our estimations, while Section 4 provides some concluding remarks.

2.   Materials and Methods

For our analyses, we make use of data from the Survey of Health, Ageing and Retirement in Europe, collected during the pandemic (SHARE Corona Survey). SHARE is a panel database of microdata on health, socio-economic status, and social and family networks, collecting information from all continental EU countries plus Switzerland and Israel. The target population consists of all the people with an age of 50 years or over at the time of sampling and with a regular domicile in a country surveyed by SHARE. For each respondent, current partners living in the same household are also interviewed, regardless of their age.

The SHARE project, started in 2004, had been collecting data for eight waves to date, providing unique information in a time in which Europe is faced with an increasingly ageing population. SHARE data collection usually relies on computer-assisted personal interviewing (CAPI). However, the COVID-19 pandemic broke out while interviews for SHARE's 8th wave were underway, making it impossible to resume fieldwork as of March 2020, when about one third of the expected interviews still had to be conducted. Therefore, in order not to suspend the activities, SHARE switched to computer-assisted telephone interviewing (CATI), developing an ad hoc questionnaire covering the same topics as the regular survey, although substantially shortened and aimed at also capturing the changed circumstances affecting respondents after the COVID-19 outbreak (Scherpenzeel et al., 2020): thus, the SHARE Corona Survey covers the most relevant life domains of individuals aged 50 or older and includes brand new questions about infections and effects of the lockdown on the respondents' daily lives.

Our research hypotheses, introduced in the previous Section, are corroborated by a number of descriptive analyses, shown in the following tables (Tables 1–5). These analyses are aimed at describing the study sample, without claiming to draw conclusions about the population.

The respondents were asked to think of their household's total monthly income since the COVID-19 outbreak and to rate their ability to meet their expenses. As Table 1 shows, the share of people declaring to receive financial support is higher among those in greater economic difficulties and it decreases with increasing ability to make ends meet, in line with H4.

Further evidence in support of H4 is given in Table 2, which reports the share of people classified by their ability to make ends meet, in the Corona Survey and in Wave 7, which took place in 2017 (Börsch-Supan, 2020), to show how the proportions changed in between. It looks like more people had difficulties in the pre-pandemic period, compared to the information collected during the pandemic: indeed, this may be due to the financial support received by the families suffering from economic difficulties, which increased their ability to make ends meet despite the general economic damages brought about by the pandemic. This is corroborated by Eurostat estimates, showing that, while the median individual employment income significantly fell in 2020 compared to 2019, median household equivalised disposable income increased; however, many countries experienced a rise in the proportion of working-age citizens at risk of poverty in 2020 compared to 2019, while only Estonia, among European Union countries, experienced a decrease of such rate (Eurostat, 2021b).

As Table 3 shows, the share of the population receiving financial aid is much higher in the East, compared to the West. Such support could come from employers, government, relatives, friends or others.

As Table 4 shows, among those who received financial aid, the great majority declares that it came from the government.

As hypothesised (H2), we would expect a direct relationship between frequency of social contacts and stated health level. Table 5 highlights such a relationship in the sample: with poorer health, the share of people declaring to never meet their acquaintances increases.

To answer our research questions, we proceed with the estimation of three models based on data from the SHARE Corona Survey: an ordinal logistic regression of households' ability to make ends meet since the outbreak (Model 1), a logistic regression of whether the respondents lost their employment due to the outbreak (Model 2), and a logistic regression of whether the respondents received financial support due to the outbreak (Model 3).

The ordinal logit model (Model 1) can be derived from a measurement model in which a continuous latent variable, ${y}^{{*}}$, is mapped to an observed variable y. The continuous unobservable propensity (${y}_{i}^{{*}}$, latent variable) would cross thresholds ($\tau$) that differentiate adjacent levels of the observed ordered ${y}_{i}$'s. The latent variable is supposed to be linearly related to the observed $x$'s through the following structural model:

where $\beta$ is the vector of coefficients and ${\varepsilon }_{i}$ is the error term with mean zero and standard deviation $\pi /\sqrt{3}$. The manifest ordinal variable ${y}_{i}$ is related to ${y}_{i}^{{*}}$ according to the following model:

where m = 1 to J identifies the number of levels of the manifest ordinal variable.

The standard logit model, used in Models 2 and 3, is:

where ${y}_{i}$ is the dependent variable; on the right side of the equation, we have ${x}_{i}$, representing the explanatory variables with coefficients ${\rm{ \mathsf{ β} }}$, and the error term ${e}_{i}$. The coefficients of the model (i.e., the ${\rm{ \mathsf{ β} }}$ parameters in the equation) are estimated by maximising the log-likelihood function.

All models employ contact frequency with neighbours, friends or colleagues since the outbreak (expressed through a 5-point Likert scale), sex, age, stated health level before the outbreak (5-point Likert scale), country's GDP growth in the second quarter of 2020 (collected from Eurostat, 2021c)¹, excess mortality in the country for the month of July 2020 (collected from Eurostat, 2021c)², and a dichotomic variable indicating whether a larger share of the country's population aged from 18 to 64 became at risk of poverty in 2020 compared with 2019, taking value 1 for yes and 0 for no (collected from Eurostat, 2021b). In addition, the three models include a country group dummy based on the United Nations Regional Groups classification (United Nations, 2021) to control for the East-West dichotomy (1: Eastern European Group; 2: Western European and Others Group).

¹The GDP growth indicator is expressed as a percentage change of the second quarter of 2020 compared with the previous quarter and as an index with base year 2015, with seasonally and calendar adjusted data.

²The excess mortality indicator is computed as the relative difference of the count of monthly deaths in July 2020 from its average for the same month over the period 2016–2019. Monthly data are estimated from weekly deaths data. Data are neither seasonally nor calendar adjusted. The month of July 2020 was chosen as a reference as it matches the SHARE Corona Survey data collection period in most of the countries in the sample.

Moreover, Model 1 includes the overall monthly income before the outbreak, a categorical variable indicating the employment status (0: not employed before the outbreak; 1: employment lost after the outbreak; 2: employed both before and after the outbreak), a dichotomic variable indicating whether financial support was received due to the outbreak, the number of people in the household, and a dichotomic variable indicating whether or not the respondent's partner is a member of the household.

Model 2 incorporates, along with the general covariates, a dichotomic variable indicating whether financial support was received by the respondent due to the outbreak.

Model 3 is estimated by considering the common variables together with the household's ability to make ends meet since the outbreak (expressed on a 4-point Likert scale), the overall monthly income before the outbreak, the employment status (0: not employed before the outbreak; 1: employment lost after the outbreak; 2: employed both before and after the outbreak), the number of people in the household, and a dichotomic variable indicating whether the respondent's partner is a member of the household.

Our regression models are unweighted. Indeed, albeit population weights are available in the SHARE dataset, it is always important to think carefully about whether the reason for using them really applies. In the present case, as population weights are a function of sociodemographic characteristics that are already included in our regression models, over- or under-representation of some population groups is already controlled for. Therefore, weighting looks to be unnecessary for consistency and could even induce heteroskedasticity, leading to imprecise (i.e., less efficient) estimates with inflated standard errors (Dickens, 1990; Winship and Radbill, 1994).

The full list of countries in the analysed sample is presented in Table 6, along with the United Nations Regional Groups classification (United Nations, 2021), according to which the sampled countries are divided into Eastern European Group (EEG) and Western European and Others Group (WEOG), as well as information on GDP growth in the second quarter of 2020 (Eurostat, 2021c), excess mortality in the country for the month of July 2020 (Eurostat, 2021c), and on whether a larger share of the country's population aged from 18 to 64 became at risk of poverty in 2020 compared with 2019 (Eurostat, 2021b).

³Cyprus is part of the Asia and the Pacific Group in the United Nations' classification. For our analyses, we include the country in the Western European and Others Group.

Table 7 shows the descriptive statistics (observations, quartiles, mean, and standard deviation) of the whole set of variables employed in our analyses.

3.   Results

Table 8 shows the results from our three models.

As mentioned in the previous Section, Model 1 is an ordinal logistic regression of households' ability to make ends meet since the outbreak. All the regressors show a very high statistical significance.

Respondents stating that they have been engaging with neighbours, friends or colleagues every day since the outbreak of COVID-19, compared to those who have not, are more likely to be able to cope with the unfavourable economic conditions brought about by the pandemic, which is consistent with our first hypothesis (H1). Indeed, compared to the respondents who engage with their neighbours, friends or colleagues on a daily basis, those declaring a contact frequency of several times a week are significantly less likely to make ends meet; those who meet their acquaintances about once a week are even less likely; and those who keep in touch less often than weekly, if ever, are even much less likely.

Needless to say, a higher monthly income before the outbreak makes respondents more likely to maintain the ability to eke out a living.

In line with our third hypothesis (H3), the respondents who lost their employment due to the consequences of the pandemic are less likely to be able to make ends meet, compared to those who were not employed even before the outbreak; conversely, those who were employed before the outbreak, and have kept their jobs since, are more likely to get through the month.

Receiving financial support due to the outbreak is associated with a lower ability to make ends meet, agreeably indicating that such support goes mainly to those who need it the most, in line with our fourth hypothesis (H4).

Males, as well as older people, are more likely to be able to make ends meet.

The lower the perceived health level, the lower the likelihood of comfortably getting to the end of the month.

A larger number of members in the household is associated with a lower likelihood of being able to cover expenditure, while the presence of a partner makes it more likely to be able to make ends meet.

Respondents from countries exhibiting a higher GDP growth, or a lower decline in GDP, in the second quarter of 2020 with respect to the previous quarter are more likely to be able to meet their expenses (H5), as well as people from countries belonging to the Western European and Others Group.

Conversely, people from countries in which the strength of the pandemic was higher, proxied by excess mortality in July 2020 compared with average monthly deaths in 2016–2019—are less likely to be able to get through the end of the month. The same goes for respondents living in countries in which a larger share of the working-age population became at risk of poverty in 2020 compared with 2019.

As the relationship between employment status and households' ability to make ends meet (H3) turned out to be particularly relevant in Model 1, we also estimate a logistic regression of whether the respondents lost their employment due to the outbreak (Model 2).

A stronger connection with acquaintances since the outbreak is negatively associated with the likelihood of respondents having lost their job: the lower the contact frequency, the higher the likelihood of reporting employment loss.

Having received financial support due to the outbreak is associated with a higher likelihood of respondents having lost their job. Therefore, also in this case, financial support seems to go to those who need it the most (H4).

Male respondents are less likely than females to report job loss. Older individuals are also more likely to report job loss, maybe as it is easier for firms to manage an early retirement for them.

Those who perceive their health level to be fair or poor, compared to those who state to be in excellent health status, are more likely to lose their jobs.

Respondents from countries characterised by a higher GDP growth, or a lower decline in GDP, in the second quarter of 2020 compared to the previous quarter are less likely to lose their jobs, in line with H5.

Conversely, respondents from countries in which the share of the working-age population at risk of poverty has increased in 2020 are more likely to lose their jobs. The same goes for people from the western part of Europe, characterised by a higher number of firms, with more potential for employment loss, and in which the SHARE sample has not received as much financial aid as in the East (see Table 3).

Finally, quite unexpectedly, the strength of the pandemic does not seem to have an effect on the probability of reporting job loss.

The last column of Table 8 shows the results from Model 3, a logistic regression of whether the respondents received financial support due to the outbreak, a variable that turned out to be particularly relevant in the previous two models.

Respondents who declare that they have been able to make ends meet since the outbreak are less likely to receive support: the easier they can get through the month, the less likely it is that they receive support. Indeed, as in the previous models, this shows that support goes to those who need it the most (H4). This is furtherly corroborated by the sign of the regressor indicating job loss: those who lost their employment after the outbreak are also more likely to receive financial support. Similarly, respondents from countries in which a higher share of the working-age population became at risk of poverty are more likely to receive support.

Those who state to never meet their acquaintances are more likely to receive support. The respondents characterised by a higher monthly income before the outbreak are also more likely to get financial aid.

Older adults are less likely to receive financial support, as we could expect by looking at the results from the previous models. Indeed, most older adults are retired and, compared to the younger respondents in the sample, did not plausibly experience a significant decrease in available income due to the outbreak, such as to be requiring financial aid. Moreover, older people might be able to tap into a higher monetary wealth, making them more capable than younger individuals to absorb income shocks. However, this may not be true for all, and policymakers should ensure that older individuals who got financially affected by the pandemic are adequately assisted.

A larger household size increases the likelihood of receiving financial support.

In this case, perceived health level, sex and presence of a partner in the household do not seem to have an effect on the probability of receiving financial support.

Living in a western European country, as well as in a country with higher GDP growth, makes it less likely to receive financial aid, in line with H5.

Finally, respondents from countries in which the pandemic strength is higher are less likely to receive financial aid: indeed, if the pandemic situation is more dramatic, we can assume that a higher number of people will experience financial distress, in a general setting in which States are already going into massive debt and are not able to cope with all the aid requests.

4.   Discussion

In this paper, we examine the economic consequences of COVID-19 on the older European population, focusing on their ability to make ends meet, on the eventual job loss recorded at the end of the first COVID-19 wave, and on the financial aid received from the government, employer, relatives and/or friends.

Our results confirm that social networks (family and friends) play a relevant role in allowing citizens to cope with the adverse economic conditions brought about by the pandemic, highlighting an interesting social component of poverty. Moreover, the frequency of social contacts reveals a positive association with health level. The citizens who lost their employment due to the pandemic are less likely to be able to make ends meet, compared to those who were not employed even before the outbreak. People who received financial support seem to be the ones who most need it. The ability to get through the month and the likelihood of job loss is positively correlated with increasing age, while older people are less likely to receive financial support, thus resulting to be less economically vulnerable than we would have expected. Those who perceive their health level to be fair or poor, compared to those who state to be healthier, are more likely to lose their jobs and less likely to make ends meet.

Our findings also come with some interesting country group differences. Respondents from countries characterised by a higher GDP growth, or a lower decline in GDP, in the second quarter of 2020 with respect to the previous quarter are less likely to lose their jobs, more likely to be able to meet their expenses, and less likely to receive financial aid. People from the western part of Europe are more likely to lose their jobs, more likely to be able to meet their expenses, and less likely to receive financial support. Respondents from countries in which the share of the working-age population at risk of poverty has increased in 2020 are more likely to lose their jobs, more likely to receive economic support, and less likely to be able to get through the end of the month. People from countries characterised by worse pandemic health conditions are less likely to be able to make ends meet and less likely to receive financial aid: if the pandemic situation is more dramatic, in a general setting in which States are already going into enormous debt, governments will not be able to cope with the aid requests received by the large share of people experiencing financial distress.

The main limitations of this study reside in the way some of the variables were measured, which warrants a cautious interpretation of the results. In particular, we must consider the subjective nature of the variable rating households' ability to make their ends meet and the absence of information regarding the amount of financial support received from the State, employer, relatives and/or friends. However, the presence of subjective perceptions in the data can also be taken as a strength, as these reflect the extent to which people are able to achieve an adequate standard of living according to their subjective needs.

Further waves of the SHARE Corona Survey will allow us to identify individual fixed effects for the analysed individuals, in order to assess whether the presented results concerning the consequences of COVID-19 on the older European population represent longer-term trends. Moreover, as the estimated models include macro-variables (GDP growth, excess mortality, risk of poverty), it would be interesting to test our hypotheses by means of multilevel models.

Acknowledgments

This paper uses data from SHARE Wave 7 and SHARE Corona Survey (DOIs: 10.6103/SHARE.w7.711, 10.6103/SHARE.w8ca.100). See Börsch-Supan et al. (2013) for methodological details. The SHARE data collection has been funded by the European Commission, DG RTD through FP5 (QLK6-CT-2001-00360), FP6 (SHARE-I3: RII-CT-2006-062193, COMPARE: CIT5-CT-2005-028857, SHARELIFE: CIT4-CT-2006-028812), FP7 (SHARE-PREP: GA N 211909, SHARE-LEAP: GA N 227822, SHARE M4: GA N 261982, DASISH: GA N 283646) and Horizon 2020 (SHARE-DEV3: GA N 676536, SHARE-COHESION: GA N 870628, SERISS: GA N 654221, SSHOC: GA N 823782) and by DG Employment, Social Affairs & Inclusion through VS 2015/0195, VS 2016/0135, VS 2018/0285, VS 2019/0332, and VS 2020/0313. Additional funding from the German Ministry of Education and Research, the Max Planck Society for the Advancement of Science, the U.S. National Institute on Aging (U01_AG09740-13S2, P01_AG005842, P01_AG08291, P30_AG12815, R21_AG025169, Y1-AG-4553-01, IAG_BSR06-11, OGHA_04-064, HHSN271201300071C, RAG052527A) and from various national funding sources is gratefully acknowledged (see www.share-project.org).

Conflict of interest

All authors declare no conflicts of interest in this paper.