Rightly blamed the ‘bad guy’? Grandparental childcare and COVID-19

View author details View Less
  • 1 German Youth Institute, , Germany
Full Access
Get eTOC alerts
Rights and permissions Cite this article

This study explores the link between regular grandparental childcare and SARS-CoV-2 infection rates at the level of German counties. In our analysis, we suggest that a region’s infection rates are shaped by region-, household- and individual-specific parameters. We extensively draw on the latter, exploring the intra- and extra-familial mechanisms fuelling individual contact frequency to test the potential role of regular grandparental childcare in explaining overall infection rates. We combine aggregate survey data with local administrative data for German counties and find a positive correlation between the frequency of regular grandparental childcare and local SARS-CoV-2 infection rates. However, the statistical significance of this relationship breaks down as soon as potentially confounding factors, in particular, the local Catholic population share, are controlled for. Our findings do not provide valid support for a significant role of grandparental childcare in driving SARS-CoV-2 infections, but rather suggest that the frequency of extra-familial contacts driven by religious communities might be a more relevant channel in this context. Our results cast doubt on simplistic narratives postulating a link between intergenerational contacts and infection rates.

Abstract

This study explores the link between regular grandparental childcare and SARS-CoV-2 infection rates at the level of German counties. In our analysis, we suggest that a region’s infection rates are shaped by region-, household- and individual-specific parameters. We extensively draw on the latter, exploring the intra- and extra-familial mechanisms fuelling individual contact frequency to test the potential role of regular grandparental childcare in explaining overall infection rates. We combine aggregate survey data with local administrative data for German counties and find a positive correlation between the frequency of regular grandparental childcare and local SARS-CoV-2 infection rates. However, the statistical significance of this relationship breaks down as soon as potentially confounding factors, in particular, the local Catholic population share, are controlled for. Our findings do not provide valid support for a significant role of grandparental childcare in driving SARS-CoV-2 infections, but rather suggest that the frequency of extra-familial contacts driven by religious communities might be a more relevant channel in this context. Our results cast doubt on simplistic narratives postulating a link between intergenerational contacts and infection rates.

Introduction

Since the beginning of the pandemic, many studies have analysed the driving forces behind the spread of SARS-CoV-2. This includes the role of social contacts for the prevalence of SARS-CoV-2 and the disease COVID-19 caused by the virus. In this context, the links between contacts within the family, potentially increased risks of infection and COVID-19 mortality have been investigated by, for example, Aparicio and Grossbard (2020), Arpino et al (2020), Balbo et al (2020) and Bayer and Kuhn (2020). In our analysis, we suggest that a region’s infection rates are mainly shaped by the two region-specific parameters of infection path and spatial distance, and the two individual-specific parameters of vulnerability and contact frequency. We extensively draw on the latter, exploring the intra- and extra-familial mechanisms fuelling contact frequency to test the potential role of regular grandparental childcare (GPC) in explaining overall infection rates. We study these relationships in Germany, combining aggregate survey data with local administrative data, and find a positive correlation between the frequency of GPC and local SARS-CoV-2 infection rates. However, the statistical significance of this relationship breaks down as soon as potentially confounding factors, in particular, the local Catholic population share, are controlled for. Our findings suggest that the frequency of extra-familial contacts driven by religious communities might be a more relevant channel of SARS-CoV-2 infections than GPC.

Due to substantially higher mortality rates of the older persons infected with SARS-CoV-2, early studies in 2020 already pointed to the vulnerability of certain regions due to their demographic characteristics (Kashnitsky and Aburto, 2020), as well as the prevalence of intergenerational relations (Balbo et al, 2020). Aparicio and Grossbard (2020) present evidence that the frequency of intergenerational co-residence in US states is positively related to COVID-19 fatalities per capita. Similar results are found by Bayer and Kuhn (2020) in a cross-country analysis with 24 countries. However, Arpino et al (2020) cannot confirm these findings. They provide a comprehensive analysis of aggregated data on intergenerational family relations from the Survey of Health, Ageing and Retirement in Europe (SHARE), linked with information on registered SARS-CoV-2 test data and case fatality rates as published by national health agencies in several European countries. They do not find a robust relationship between infections or case fatality rates and their key variables of interest on the family level, including the frequency of intergenerational contacts, the share of intergenerational households and the prevalence of GPC in a region or country.1

With data for Germany, a comprehensive analysis of the potential link between the extent of GPC support and SARS-CoV-2 infections has not yet been conducted. The SHARE data used by Arpino et al (2020), for example, do not allow for an analysis on the fine-grained local level because location information of respondents is only made available on the level of federal states. Moreover, previous studies have not convincingly investigated the role of potentially confounding factors when analysing the correlation between GPC support and SARS-CoV-2 infections. Our study aims to fill these gaps and contribute to the existing literature with an analysis for the case of Germany.

We study a potential relationship between GPC support and SARS-CoV-2 infection rates in Germany, combining survey data and registered infections at the level of German local administrative units (Kreise [‘counties’]). We draw on a comprehensive register of SARS-CoV-2 infections registered by the Gesundheitsämter (‘local German health authorities’) since the beginning of the pandemic and link these data with aggregated survey data on GPC support at the local regional level. We also provide additional micro-foundations for our analysis regarding intergenerational support based on rich individual-level survey data.

Theoretical considerations and hypotheses

In our analysis at the county level, we restrict attention to registered SARS-CoV-2 infections among those aged 60 years or older. We calculate infection rates in each county for the registered population aged 60 years or older. The reason for the age restriction is that we are interested in whether childcare responsibilities lead to higher risk of infections among the older, grandparental population. In our main analysis, we use 23 March 2020 as a reference point when counting all infections among those aged 60 years or older in the respective county.2 The date of 23 March was when the first official policy restrictions were announced by the federal government in Germany as a response to the accelerating spread of SARS-CoV-2.

We theorise that beyond individual vulnerability in terms of health status, which we capture with the individual’s age-group affiliation, three mechanisms might have shaped an individual’s exposure to the SARS-CoV-2 virus on 23 March 2020 and may therefore have driven the number of infections in the population aged 60 and over at the county level at that time (see Figure 1). The first mechanism refers to the infection path since, due to the high infectiousness of the SARS-CoV-2 virus, the moment in time when the first infection is measured takes the respective county to a higher level of virus dissemination, measured by the increase in infection numbers (for example, Bouffanais and Lim, 2020; Frieden and Lee, 2020). Second, the spatial distance of people, captured by the settlement structure, moderates the infection path such that, other things being equal, densely populated areas will reinforce, and sparsely populated areas will decrease, the dissemination path of the virus. The mechanism behind settlement structure is that spatial distance to other human beings is more easily kept in more spacious areas (for example, Rader et al, 2020). Third, and of utmost importance to this article, the frequency of social contacts, which can be subdivided into intra-familial and extra-familial contacts, is decisive. Although this mechanism is mutually linked to settlement structure and infection path, we argue that it is by itself a function of individuals’ daily lives, which are shaped by individual and household context-related characteristics, as well as macro-level norms, institutions and economic activity. Concerning the extra-familial sphere, the job context as well as non-job activities should shape an individual’s interpersonal interactions. Inside the family, interactions refer to individuals of the same generation (intragenerational), as well as to upward intergenerational (the generation of children that might be engaged in eldercare for their parents) and downward intergenerational (the generation of grandchildren that might be subject to GPC) interactions. In this article, it is this last type of intra-familial contacts that we are interested in.

Figure 1:
Figure 1:

Illustration of causal mechanisms explaining infection rates at the county level

Citation: Journal of Public Finance and Public Choice 2022; 10.1332/251569121X16152354192487

Source: Own illustration.

Since GPC is only one source of social contacts, our hypotheses, which structure our empirical investigation, will focus on potential drivers of GPC and, at the same time, on extra-familial sources of social contacts that might be confounding variables when analysing the link between GPC and the observed overall infection rates in the elderly population on the county level. We argue that GPC should be shaped by the household context, institutional childcare facilities available on the local level as a substitute for within-family care (H1) and norms, that is, reciprocity norms and the value attributed to intergenerational relations (H2).

In more detail, we rely on the extensive margin of regular GPC in our main analysis, exploiting the information on the use of regular childcare. Starting with the household context (H1), we expect that working parents, especially working single mothers, should increase the likelihood of GPC since grandparents could compensate for scarce parental time resources in this case (Hank and Buber, 2009). Compared to small children under three for which substitutes for parental care are more difficult to find, the presence of a (youngest) child aged three to five should be related to more GPC (Hank and Buber, 2009). However, GPC should become less frequent the higher the number of children is (Jappens and van Bavel, 2012). Institutional childcare usage at the individual level should decrease the need for GPC (Albertini and Kohli, 2013). Further, concerning norms, we expect that Catholic denomination is related to stronger intergenerational family ties and should therefore increase the propensity of GPC (H2). A central foundation of Catholic social thought is the so-called subsidiary concept, emphasising the role of intra-familial solidarity for social support (for example, Gundlach, 1964; Althammer, 2013). Thus, we expect Catholic denomination to be positively associated with the prevalence of social norms fostering within-family support, for example, GPC.

Regarding infection rates at the county level, we postulate that the infection path, in terms of (log) days since the first patient, increases the number of infections since the virus has had more time to disseminate (H3). We further hypothesise that infection rates increase with population density, measured in four settlement types (H4). Moreover, we suggest that children enrolled in public childcare, as well as parents engaged in job-related social contacts, could provide a source of infection for the (grand)parents they interact with. Therefore, we use institutional childcare coverage rates for children below age three and aged three to five, respectively (H5), as well as (log) median income as a proxy for economic activity (H6), as explanatory variables in the infections equation.3 We thereby rely on previous studies suggesting a positive linkage between economic activity and both infections (for example, Adda, 2016; Rader et al, 2020) and institutional childcare, respectively. The intuition behind income is that a higher economic added value is related to more job-related contacts. Remote work, especially work from home, has been shown to be more likely offered to and used by the highly educated workforce (Alipour et al, 2020) at the top of the earnings distribution. Since remote work was not as prevalent in March 2020 as it is today, this counter-effect was probably not strong enough to outweigh the opposite (sales- and revenues-driven) positive linkage between income and infections.

Due to a higher risk exposure of elderly people who take care of their grandchildren, we expect that regular GPC should be associated with higher infection rates (H7). However, we doubt that GPC is the true source of this phenomenon. Rather, we suggest that influential third variables drive the association between GPC and infections at the county level. We thereby follow Arpino et al (2020), who argue that a stronger focus on within-family ties and a correspondingly lower weight of extra-familial ties could serve as a shield protecting the elderly against the virus. As pointed out by the authors, elderly people with close relationships to their children and grandchildren might rely less on social contacts outside the family, which might potentially involve even bigger threats of infections. Second, strong intergenerational relationships might affect family life: members might be more careful regarding social interactions outside the family; and, in some of these cases, grandparents might also live close to their children and grandchildren, or even in the same house. Third, there is also evidence that family ties and interactions can have a positive effect on psychological well-being and health, decreasing the risk of an infection (see, for example, Cohen, 2020). In sum, these arguments would motivate a negative association between regular GPC and infection rates.

Beyond the indirect channel via GPC, Catholic denomination might impact infections directly via a networks mechanism related to the private (non-job) sphere. Religious activities and other ritual activities do not occur unless a sufficient number of followers is reached at the local level. In particular, religious events, for example, worship and religiously motivated celebrations with a high number of participants, such as weddings, are suspected to drive infection numbers (Lee et al, 2020; Salvador et al, 2020). Moreover, though not directly associated with religiosity in contemporary Germany, German Rhineland regions, for example, Rhineland-Palatinate and North Rhine-Westphalia, where many Catholic people live, are well known for their carnival processions, which take place in February. Thus, via the regional bracket, counties with a high population share of Catholics might exhibit higher infection numbers via the non-job networks channel (H8). An extensive literature has studied the role of religiosity for extra-familial social networks (for Germany, see, for example, Traunmüller, 2009).

Moreover, due to a higher contact frequency among younger people, a high population share of minors and a low share of elderly people (aged 60 or over) should increase infection rates. In sum, population composition by age should play a role too (H9). As a further control variable, we include the share of foreign nationals per county, as well as a dummy variable for Eastern German counties that belonged to the former German Democratic Republic (GDR).

Empirical analysis

Data and descriptive statistics

For our subsequent analysis in the light of the hypothesised mechanisms, we combine different data sources. We draw on individual-level survey data from the sixth wave of Kinderbetreuungsstudie (‘child care study’ KiBS), which is administered by the German Youth Institute (Aust et al, 2018; Alt et al, 2020). The data were collected in 2017 and involved 36,800 interviews conducted among Auskunftspersonen (‘reference persons’) of children in the target population below the age of 15 living in 249 selected counties in Germany. We apply survey weights as described in Alt et al (2020), correcting for non-response bias based on a two-stage weighting procedure. First, observations are weighted according to administrative statistical information on the distribution of children according to age groups in the residence state. Second, the weighting also accounts for the non-response behaviour of reference persons according to different institutional childcare arrangements. However, our main results are not sensitive to using the weighted or the unweighted sample in our analysis. The variable of interest in our analysis is the indicator of whether grandparents provide regular childcare to their grandchildren. KiBS contains information on whether grandparents provide regular childcare support for a child. The answer options in the survey questionnaire were ‘regular use’, ‘as required’ and ‘not at all’.4 If a respondent chose ‘regular use’, they were asked for how many hours this support was provided in a regular week. In our analysis, we use the survey information on whether a child receives regular GPC as a binary variable at the individual level, as well as in the form of weighted respondent shares at the county level. For our micro-founded analyses, we draw on further individual-level information from the KiBS survey, such as the child’s age, the household composition and labour force participation.

Second, we use county-level data on SARS-CoV-2 infections in Germany, as collected by the local health authorities (Gesundheitsämter) and published by the Robert-Koch Institute (RKI, 2020). As central information from these data, we use case counts per 100,000 inhabitants at the county level (at different points in time). In our main analysis, we use infection register data from before the first policy restrictions were announced on 23 March 2020 to circumvent any differential effect of these restrictions that could be correlated with our variables of interest. In addition to the total recorded SARS-CoV-2 cases, the RKI has also published corresponding figures by age group. From the RKI data, we additionally calculate the time that has passed since the first SARS-CoV-2 case recorded in a given county. Third, we use administrative data on sociodemographic and economic characteristics at the county level, as provided by the interactive database Indikatoren und Karten zur Raum- und Stadtentwicklung (INKAR) of the Bundesinstitut für Bau-, Stadt- und Raumforschung (BBSR, 2020).

In Appendix A, we present some descriptive statistics for the key variables of our analysis (for detailed information, see Tables A1A4 in Appendix A). Table A1 depicts the shares of GPC aggregated from the KiBS data for the 16 federal states and four county types, distinguishing rural and urban counties. The data show that GPC is less common in northern than in southern states in Germany and also less common in the so-called city-states of Hamburg, Berlin and Bremen. This holds for all regular GPC, as well as for GPC that covers more than seven hours per week. Table A2 reports these shares separately by children’s age groups (0–3 versus 3–5 versus 6–14), revealing that regular GPC support is most frequent in the 3–5 age group. Table A2 additionally shows the share of intense regular GPC (more than seven hours per week) per child age group. However, we argue that it is not intensity, but regularity, that is decisive for infection transmission on the micro level, and a significant share of families using GPC is required to establish a related link between GPC and infection rates on the macro (county) level. We therefore adhere to our sample specification of families with 0–14-year-old children, focusing on regular GPC in our main analysis and intense GPC in our sensitivity analysis.

As mentioned earlier, the use of institutional childcare arrangements in a county might be an important factor in this context if grandparents step in when no other childcare possibilities are available. Column 4 in Table A1 shows that institutional childcare coverage for children between age three and five is generally very high (> 89 per cent) in all states and county types. In contrast, it is considerably lower for children below the age of three (see column 3 in Table A1). The coverage in Städtischer Kreis (‘cities’) tends to be lower than in rural areas and Kreisfreie Großstadt (‘large metropolitan areas’), but the variation between different county types is not as large as between federal states. In general, coverage rates for the under-threes are higher in Eastern compared to Western Germany.

Table A3 denotes total infection numbers and numbers per 1,000 inhabitants by German federal states for 23 March 2020. Additional to our focus group of elderly people aged 60 and over, figures for the total population are presented. Concerning the elderly, infections per 1,000 inhabitants vary between 24.2 (Mecklenburg-Western Pomerania) to 185.6 (Baden-Wuerttemberg). With respect to the total population, cases range from 81 in Bremen to 4,723 in Bavaria. Table A4 depicts total infection numbers and numbers per 1,000 inhabitants by German federal states for 30 September 2020, which is used as an alternative reference point in our robustness checks.

Main regression analysis

Obviously, GPC is only a small piece of the puzzle explaining overall infection rates in the target population of elderly people. Therefore, our empirical strategy consists of two strands. First, we build on the theoretical underpinnings of GPC. This micro-foundation will undergo an empirical test relying on the KiBS data outlined earlier. Via its aggregated form on the county level – the population share using regular GPC – the outcome variable on the micro level will enter the infections equation on the macro (county) level as the second strand of our empirical design.

Starting with the micro-foundation of regular GPC, Table 1 presents the results of linear probability regressions with the weighted sample of 33,259 children in the target population of those aged 0 to 14 in Germany reached by the KiBS survey. The dependent variable is whether a child in the sample is regularly taken care of by a grandparent. Indeed, the results in Table 1 document that the population share of Catholics on the county level is strongly positively associated with the probability that the grandparents are involved in childcare support, confirming our hypothesis H2. Depending on the specification, this roughly means that a 10 per cent increase in the Catholic population share in a county is associated with a 0.4 to 0.6 percentage-point increase in the probability that the grandparents of a child are regularly involved in childcare.

Table 1:

Analysis of micro-level characteristics associated with regular GPC support

Catholic population share0.0619***0.0406***0.0528***
(0.0104)(0.0107)(0.0172)
Child age 0–2 (reference category)
Child age 3–50.0314***0.0276***0.0276***
(0.00736)(0.00736)(0.00737)
Child age 6–10–0.00776–0.0140**–0.0143**
(0.00661)(0.00664)(0.00665)
Child age 10–14–0.127***–0.132***–0.132***
(0.00675)(0.00677)(0.00677)
Single mother, working0.129***0.129***0.128***
(0.0168)(0.0168)(0.0168)
Single mother, not working0.0761***0.0714**0.0705**
(0.0289)(0.0289)(0.0289)
Single father, working0.103*0.105*0.109**
(0.0537)(0.0536)(0.0536)
Single father, not working–0.0381–0.0158–0.00985
(0.113)(0.113)(0.113)
Couple, male breadwinner (reference category)
Couple, female breadwinner0.0316**0.0340**0.0342**
(0.0143)(0.0143)(0.0143)
Couple, dual earner0.131***0.131***0.130***
(0.00555)(0.00555)(0.00555)
Couple, not working0.002140.004390.00501
(0.0154)(0.0153)(0.0154)
1 child in household (reference category)
2 children in household–0.0161***–0.0164***–0.0162***
(0.00546)(0.00545)(0.00546)
3+ children in household–0.0552***–0.0547***–0.0547***
(0.00641)(0.00640)(0.00641)
Institutional childcare attendance–0.0460***–0.0417***–0.0416***
(0.00503)(0.00505)(0.00508)
Rural county – sparsely populated (reference category)
Rural county0.0176**0.0164*
(0.00850)(0.00885)
Urban county0.003860.00394
(0.00752)(0.00877)
Urban county (single metropolitan area)–0.0330***–0.0357***
(0.00757)(0.00848)
State fixed effectsNoNoYes
Observations33,25933,25933,259

Notes: Weighted ordinary least squares (OLS) regressions (linear probability model). *** p < 0.01; ** p < 0.05; * p < 0.1, standard errors in parentheses.

Moreover, these micro-level results reveal that children between the ages of three and five are more frequently taken care of by their grandparents; children aged six to ten, and particularly those aged ten to 14, are less likely to receive regular GPC compared to the reference group of children under the age of three. As to parents’ labour force participation status, the reference category is defined as couples pursuing a traditional male-breadwinner model, that is, the male partner works and the female partner is out of the labour force. Relative to this reference group, single mothers in as well as out of the labour force are more likely to receive childcare support from the grandparents of their children. For single fathers, this is only the case if they are in the labour force (the group of single fathers in the sample contains only 60 observations). For couples that pursue a dual-earner model and couples where only the female partner participates in the labour market, it is also more likely that the grandparents of their children provide regular childcare support. This analysis shows that certain family types, for example, single mothers and dual-earner couples, rely on support by grandparents relatively more often than traditional male-breadwinner families. In case of an available and used institutional childcare arrangement, grandparents are less likely to be involved in regular childcare support. In sum, the household context variables meet our expectations (H1).

In column 2, we additionally include information on the county settlement structure (rural or urban), in column 3 we include federal state dummies. Our previous findings are robust to the inclusion of these additional variables.

We now turn to the question of whether the presented data reveal any relationship between regular GPC support and higher rates of SARS-CoV-2 infections in a county. Table 2 reports results of ordinary least squares (OLS) regression estimations at the county level, linking aggregated information on regular GPC support for children aged under 15 from KiBS at the county level5 with regional administrative statistics and infection rates, as introduced earlier. Using the logarithm of cumulative infection rates in a county on 23 March as the dependent variable, we subsequently introduce various regional variables to test our hypotheses and account for potentially confounding factors in the correlation analysis. We present population-size-weighted estimates. Results are qualitatively similar when running the regressions without population weights.

Table 2:

Regressions explaining log registered infections per 100,000 (60+) (23 March 2020) at the county level

Log registered infections per 100,000 (60+)
Log days since first case1.478***0.929***0.916***0.947***0.914***
(0.168)(0.186)(0.183)(0.184)(0.174)
Share of regular GPC (below age 15)1.658**1.607**1.531**1.594**0.481
(0.708)(0.683)(0.671)(0.682)(0.651)
Log population/km2–0.0587–0.123**–0.0544–0.0903
(0.0506)(0.0587)(0.0710)(0.0712)
Log median income1.747***1.273**1.442*1.176*
(0.581)(0.622)(0.751)(0.708)
Eastern Germany–0.09030.0179–0.08440.311
(0.149)(0.149)(0.220)(0.273)
Urban county0.362***0.311**0.286**
(0.131)(0.137)(0.128)
Population share age 60+–1.842–1.7481.969
(2.606)(2.767)(2.798)
Population share under 187.3719.052*10.38**
(4.698)(5.187)(5.005)
Share institutional childcare (below age 3)0.001860.00246
(0.00716)(0.00943)
Share institutional childcare (age 3–5)0.01640.0164
(0.0142)(0.0154)
Foreigner share–1.5510.478
(1.706)(1.637)
Catholic population share1.375***
(0.257)
Observations (counties)197197197197188

Notes: Only counties with at least 30 individual observations in KiBS survey data, population-weighted coefficient estimation. *** p < 0.01; ** p < 0.05; * p < 0.1, standard errors in parentheses.

Column 1 shows a statistically significant positive relationship between log SARS-CoV-2 infection rates and the percentage share of GPC in a county, supporting H7. In this specification, we only additionally account for the days since the first registered SARS-CoV-2 case in a county (which are significantly positively associated with infection rates, supporting H3). These otherwise unconditional correlation results could be interpreted as confirming the hypothesis that more frequent contact of the old-age population is associated with higher risk of infection (for example, Aparicio and Grossbard, 2020). In columns 2–5, we introduce further variables to our empirical model. Counties with higher median income and metropolitan counties exhibit higher infection rates in the old-age population, providing support for our hypotheses H4 and H6. Childcare coverage rates and the population shares of non-Germans and elderly people, respectively, lack significant associations with county-specific infection rates.

Most importantly, when including the share of Catholic population in column 5, we observe that this variable is highly positively correlated with our dependent variable (confirming H8), and that it absorbs the effect of GPC.6 Therefore, H7 has to be discarded in this specification. This finding shows that there is no valid link between GPC and infection rates on the county level as soon as further region-specific characteristics are controlled for. Therefore, regional variation in GPC cannot be seen as a driver of regional variation in infection rates. Although affiliation with the Catholic religion increases the likelihood of GPC, the Catholic effect is obviously not limited to the family sphere, but drives infection rates even if its within-family effect is accounted for. We motivate this extra-familial effect with social networks related to religious beliefs. Our regression results in Table 2 support this suggestion.

The population share of minors is significant in the second and even more so in the third specification. We do not find any effect of the population share of those aged 60 and over. Therefore, on the basis of our data, hypothesis H9 can only partially be confirmed and H5 cannot be confirmed at all.

In our next step, we build on the findings from Table 2 and particularly on the interpretation that Catholic denomination fuels infection rates via both a GPC and a social networks channel. On the one hand, following from H1, which has been confirmed in our GPC estimations, a higher share of Catholics in the population should increase GPC, which, in turn, proved to increase infection risks in our infection regressions (see columns 1–4 in Table 2). Following this reasoning, we could expect a stronger/significant relationship between regular GPC support and infection rates in these counties.

On the other hand, as discussed in the empirical strategy section, we suppose that the extra-familial channel requires a sufficient population share of Catholics to become effective: to celebrate religious events with non-family members and to establish religion-related rituals, a threshold of peers with the same attitudes is required. As mentioned earlier, due to the region commonality, we subsume carnival processions in this category. We cannot measure the non-job networks channel directly, but we suggest that in counties lacking a significant share of Catholics, the extra-familial mechanism should be relatively weaker than the within-family mechanism, attributing regular GPC a higher role in this environment. Conversely, in counties exhibiting a significant share of Catholics, we expect the regular GPC effect to be relatively weaker and the extra-familial channel to be relatively stronger. To disentangle the GPC from the Catholic extra-familial channel and to verify which argumentation is supported by our data, our next task is to test H7 again, this time in two different social environments, distinguished by the prevalence of Catholic denomination. While this factor entered Table 2 as a ‘shift effect’, it is allowed to interact with all independent variables in the following regressions. To this end, we divide our sample into two subsamples, with the first comprising counties with a population share of Catholics below 20 per cent and the second comprising the remaining counties. We selected a threshold of 20 per cent Catholic population share as this divides our sample roughly in half. Since Eastern Germany lacks representation in the first group, we restrict both subsamples to Western German counties. The first group comprising counties with only minor shares of Catholics spans counties in Northern Germany. In 2011, for example, Schleswig-Holstein exhibited a Catholic share of 6.4 per cent, while Lower Saxony stood at 18.3 per cent (Frerk, 2018a). The second group comprising counties with Catholic population shares of 20 per cent and over concentrates in Southern and Western Germany. In 2015, the population share of Catholics stood at 39.3 per cent in North Rhine-Westphalia, at 42.2 per cent in Rhineland-Palatinate, at 59.8 per cent in the Saarland, at 34.5 per cent in Baden-Wuerttemberg and at 51.2 per cent in Bavaria (Frerk, 2018b). We use the model specification presented in column 4 of Table 2 and run our OLS regressions based on the two aforementioned subsamples.

Table 3 reports the results. As can easily be seen, regular GPC is not significant in either of the two county groups. Apparently, neither in South-Western Germany nor in the rest of the country was regular GPC significantly related to infection rates in March 2020. Even in the counties where Catholics form a relevant population subgroup, regular GPC does not significantly relate to infection rates among the elderly population. If anything, the data point to the importance of extra-familial ties: the regression coefficient for regular GPC support is larger in those counties with a Catholic share of less than 20 per cent. However, the coefficients lack significance. A cautious interpretation of this finding would be that it further strengthens our interpretation derived from column 5 in Table 2: the ‘bad guy’ role of GPC is lost as soon as relevant third variables come on stage.

Table 3:

Regressions explaining log registered infections per 100,000 inhabitants (60+) (23 March 2020) at the county level (Western Germany)

Log registered infections per 100,000 (60+)
Catholic population < 20%Catholic population ≥ 20%
Log days since first case–0.0394–0.1041.185***1.098***
(0.474)(0.486)(0.229)(0.225)
Share of regular GPC (below age 15)1.6841.9530.1880.0981
(2.131)(2.182)(0.869)(0.843)
Log population/km2–0.0685–0.124–0.0648–0.0472
(0.195)(0.212)(0.106)(0.103)
Log median income3.332*3.528*1.0231.128
(1.845)(1.881)(0.960)(0.932)
Urban county0.4640.533*0.05290.194
(0.294)(0.312)(0.201)(0.203)
Population share age 60+7.0845.723–6.007–2.544
(6.591)(6.921)(4.714)(4.769)
Population share under 18–0.609–2.5243.9148.906
(13.05)(13.44)(6.468)(6.571)
Share institutional childcare (below age 3)0.03300.0305–0.0136–0.0136
(0.0234)(0.0239)(0.0135)(0.0131)
Share institutional childcare (age 3–5)–0.0396–0.02910.03610.0423*
(0.0344)(0.0377)(0.0246)(0.0239)
Foreigner share–4.019–2.164–2.908–0.735
(4.974)(5.660)(2.474)(2.546)
Catholic population share–2.5051.222**
(3.538)(0.480)
Observations (counties)40409797

Notes: Only Western German counties with at least 30 individual observations in KiBS survey data are included, population-weighted coefficient estimation. *** p < 0.01; ** p < 0.05; * p < 0.1, standard errors in parentheses.

Sensitivity analysis

We conduct two robustness checks. The first refers to our dependent variable in the GPC equation. Deviating from the standard for the main analysis defined earlier, we use the share of regular GPC for children aged 0–14 that covers more than seven hours per week to check whether GPC intensity makes a difference for our results (for the detailed results, see Table A5 in Appendix A).7 Table A5 replicates the model specifications of Table 2, the only difference lies with the specification of the GPC variable, using the intensive margin instead of the extensive margin of GPC. The main takeaway from the results is that, different from the extensive margin, regular GPC is insignificant in all specifications when the intensive margin is used: the prevalence of regular GPC of more than seven hours a week does not significantly relate to a county’s infection rates among elderly people. This is intuitive since, as mentioned earlier, for virus dissemination, the frequency of contacts should play a higher role than intensity in terms of duration.

The second robustness check concerns the reference point in time regarding registered infections. Deviating from the standard defined earlier, we base our analysis on 30 September 2020 and check the stability of our results against this change.8 For our regressions, we use the variable specifications denoted in Table 2. Table A6 and Table A7 in Appendix A provide detailed regression results. In Table A6, analogous to the results derived for 23 March 2020, the parameter of regular GPC turns insignificant as soon as the population share with Catholic denomination is taken into account. Regarding the other independent variables, results resemble those of March too: metropolitan areas and a high population share of minors are positively associated with infection numbers. Different from the March results, the population share of the non-German population is now positively related to infection rates too. Differences to the results in our main analysis could be driven by the changed regional distribution of hot spots between March and September 2020 (see Table A4 in Appendix A): during that time, the virus had been disseminating further. By the end of September, parts of Central and Northern Germany (for example, some Hessian regions and the city-state of Hamburg) exhibit rather high infection dynamics. By contrast, states in North-Eastern Germany (that is, Mecklenburg-Vorpommern, Sachsen-Anhalt, Brandenburg and Schleswig-Holstein) still feature lower infection rates. These states are characterised by rather low foreign population shares (see Table A1 in Appendix A), which might have driven the positive linkage of foreign population share to infection rates by the end of September.

Conclusion

This study explores the role of regular GPC for SARS-CoV-2 infection rates at the level of German counties. Our results show that a significant association between the two vanishes as soon as the Catholic population share is accounted for. Although Catholic religion increases the likelihood of GPC in a family, our regressions of infection rates on a range of county-related covariates show that the Catholic effect drives infection rates even if the population share using GPC effect is accounted for. We motivate this extra-familial channel with social networks related to religious beliefs. Our suggestion is tentatively confirmed by regressions based on two subsamples of counties, differing in their Catholic population share. Even in the subsample with a notable prevalence of Catholics, the GPC parameter lacks significance. However, its effect size is lower, tentatively confirming our network hypothesis assigning extra-familial mechanisms a relatively higher role in these environments.

Our main result still holds when we use 30 September 2020 instead of 23 March 2020 as a point of reference for the infection rates: here too, the significance of regular GPC is lost as soon as the population share with Catholic denomination is accounted for. By contrast, drawing on the intensive instead of the extensive margin regarding GPC does not yield any significant association between GPC prevalence and infection rates in any of the specifications. This does not contradict our main finding, but is intuitive since for virus dissemination, the frequency of contacts should play a higher role than their intensity in terms of duration.

In sum, our findings cast doubt on simplistic narratives postulating a link between intergenerational contacts and infection rates. At least, our data do not provide valid support for a significant role of GPC in driving infections; rather, our findings support previous evidence highlighting the decisive role of third variables that have to be taken into account. Exemplified with Catholic denomination, we show that region-specific third variables may enforce the postulated (intra-familial) mechanism but, at the same time, fuel extra-familial channels that limit or, in our case, even eliminate the statistical relevance of GPC. In principle, a protective effect of intra-familial ties would also be possible but our data are not suited to investigating this. Our data do not provide reliable support for a ‘bad guy’ role of GPC.

Notes

1

Dowd et al (2020) argue that results from these aggregate-level analyses should not be taken as evidence against a link between intergenerational relations and SARS-CoV-2 infection risks. However, this is not the scope of analysis in Arpino et al. (2020). Moreover, due to the lack of data on infections and the frequency/intensity of family contacts on the individual level, all studies to date rely on aggregate survey and/or administrative regional data.

2

We also conducted our analysis using infection rates in the total population as the dependent variable. Those results are slightly weaker but qualitatively similar and not statistically significantly different from the results in our main analysis. They are available upon request.

3

Regarding institutional childcare, it is debated whether children in institutional care could also be a driver of infections in the population. Recent studies argue that infected but asymptomatic children are likely to be a source of further SARS-CoV-2 contagions (Hippich et al, 2020; Laxminarayan et al, 2020).

4

For the exact wording of the question in the questionnaire, see Appendix B.

5

We only include counties with at least 30 individual-level observations. Results are qualitatively similar without the minimum number of observations restriction or when restricting to at least 20 or 50 observations per county. The average number of underlying individual-level survey observations per county in the analysis presented in Table 2 is 184. As described earlier, we apply survey non-response weights as described in Alt et al (2020).

6

When including state fixed effects in the specifications in Table 2, the coefficient for GPC support also becomes insignificant in columns 1–4. This suggests that the positive relationship between GPC support and SARS-CoV-2 infections in these columns is instead driven by the variation between states. In another extended analysis, we also included the unweighted average of infection rates in surrounding counties as an additional control variable. Here, the coefficient for GPC support also turns insignificant in columns 1–4. These estimation results are available upon request.

7

We also conducted the regressions at the micro level with the dependent variable GPC (more than seven hours per week). The results are qualitatively similar to our main analysis and available upon request.

8

In contrast to the situation before 23 March 2020, without regional differences in policy measures or testing strategies, region-specific policy responses to changing infection rates had been developed by the German states over the course of the pandemic. This is why we base our main analysis on registered infections until 23 March 2020.

Acknowledgements

We thank Paul David Boll for editorial advice and Jochen Wirsing for geodata support. Anil Eldiven and Asalia Franz provided excellent research assistance. We also thank two anonymous referees for their comments and suggestions.

Conflict of interest

The authors declare that there is no conflict of interest.

References

  • Adda, J. (2016) Economic activity and the spread of viral diseases: evidence from high frequency data, Quarterly Journal of Economics, 131(2): 891941. doi: 10.1093/qje/qjw005

    • Search Google Scholar
    • Export Citation
  • Albertini, M. and Kohli, M. (2013) The generational contract in the family: an analysis of transfer regimes in Europe, European Sociological Review, 29(4): 82840. doi: 10.1093/esr/jcs061

    • Search Google Scholar
    • Export Citation
  • Alipour, J.V., Falck, O. and Schüller, S. (2020) Homeoffice während der pandemie und die implikationen für eine Zeit nach der Krise, Ifo Schnelldienst, 73(7): 306.

    • Search Google Scholar
    • Export Citation
  • Alt, C., Anton, J., Gedon, B., Hubert, S., Hüsken, K., Lippert, K. and Schickle, V. (2020) DJI-Kinderbetreuungsreport 2019. Inanspruchnahme und Bedarf aus Elternperspektive im Bundesländervergleich, München: DJI.

    • Search Google Scholar
    • Export Citation
  • Althammer, J. (2013) Caritas in veritate: Katholische Soziallehre im Zeitalter der Globalisierung, Berlin: Duncker & Humblot.

  • Aparicio, A. and Grossbard, S. (2020) Intergenerational residence patterns and COVID-19 fatalities in the EU and the US, IZA Discussion Paper No. 13452.

    • Search Google Scholar
    • Export Citation
  • Arpino, B., Bordone, V. and Pasqualini, M. (2020) No clear association emerges between intergenerational relationships and COVID-19 fatality rates from macro-level analyses, Proceedings of the National Academy of Sciences (PNAS), 117(32): 19116121. doi: 10.1073/pnas.2008581117

    • Search Google Scholar
    • Export Citation
  • Aust, F., Burg, J.v.d., Hess, D. and Knecht, B. (2018) Methodenbericht – Kinderbetreuungsstudie 2017 und Zusatzuntersuchung ‚Pilotstudie: Räumliche Mobilität von Familien und Kindern in Zeiten der Digitalisierung, Bonn: Infas.

    • Search Google Scholar
    • Export Citation
  • Balbo, N., Billari, F.C. and Melgrano, A. (2020) The strength of family ties and COVID-19, Contexts, American Sociological Association, https://contexts.org/blog/structural-shocks-and-extreme-exposures/#balbo.

    • Search Google Scholar
    • Export Citation
  • Bayer, C. and Kuhn, M. (2020) Intergenerational ties and case fatality rates: a cross-country analysis, IZA Discussion Paper No. 13114.

  • BBSR (Bundesinstitut für Bau-, Stadt- und Raumforschung im Bundesamt für Bauwesen und Raumordnung [BBR]) (2020) Indikatoren und Karten zur Raum- und Stadtentwicklung (INKAR), Bonn: BBSR.

    • Search Google Scholar
    • Export Citation
  • Bouffanais, R. and Lim, S.S. (2020) Cities – try to predict superspreading hotspots for COVID-19, Nature, 583(7816): 3525. doi: 10.1038/d41586-020-02072-3

    • Search Google Scholar
    • Export Citation
  • Cohen, S. (2020) Psychosocial vulnerabilities to upper respiratory infectious illness: implications for susceptibility to coronavirus disease 2019 (COVID-19), Perspectives on Psychological Science, 16(1): 114.

    • Search Google Scholar
    • Export Citation
  • Dowd, J.B., Block, P., Rotondi, V. and Mills, M.C. (2020) Dangerous to claim ‘no clear association’ between intergenerational relationships and COVID-19, PNAS, 117(42): 259756. doi: 10.1073/pnas.2016831117

    • Search Google Scholar
    • Export Citation
  • Frerk, C. (2018a) Deutschland (2): Der evangelische Norden, https://fowid.de/meldung/deutschland-2-evangelischer-norden.

  • Frerk, C. (2018b) Deutschland (3): Der katholische Süden und Westen, https://fowid.de/meldung/deutschland-3-katholische-sueden-und-westen.

  • Frieden, T.R. and Lee, C.T. (2020) Identifying and interrupting superspreading events – implications for control of severe acute respiratory syndrome coronavirus 2, Emerging Infectious Diseases, 26(6): 105966. doi: 10.3201/eid2606.200495

    • Search Google Scholar
    • Export Citation
  • Gundlach, G. (1964) Die Ordnung der Menschlichen Gesellschaft (2 Bde), Köln: Bachem.

  • Hank, K. and Buber, I. (2009) Grandparents caring for their grandchildren: findings from the 2004 survey of health, ageing, and retirement in Europe, Journal of Family Issues, 30(1): 5373. doi: 10.1177/0192513X08322627

    • Search Google Scholar
    • Export Citation
  • Hippich, M. et al. (2020) Public health antibody screening indicates a six-fold higher SARS-CoV-2 exposure rate than reported cases in children, Med, 2(2): 115.

    • Search Google Scholar
    • Export Citation
  • Jappens, M. and van Bavel, J. (2012) Regional family norms and child care by grandparents in Europe, Demographic Research, 27(4): 84120. doi: 10.4054/DemRes.2012.27.4

    • Search Google Scholar
    • Export Citation
  • Kashnitsky, I. and Aburto, J.M. (2020) COVID-19 in unequally ageing European regions, World Development, 136: 105170. doi: 10.1016/j.worlddev.2020.105170

    • Search Google Scholar
    • Export Citation
  • Laxminarayan, R., Wahl, B., Dudala, S.R., Gopal, K., Neelima, S., Reddy, K.J., Radhakrishman, J. and Lewnard, J.A. (2020) Epidemiology and transmission dynamics of COVID-19 in two Indian states, Science, 370(6517): 6917. doi: 10.1126/science.abd7672

    • Search Google Scholar
    • Export Citation
  • Lee, E.C., Wada, N.I., Grabowski, M.K., Gurley, E.S. and Lessler, J. (2020) The engines of SARS-CoV-2 spread, Science, 370(6515): 4067. doi: 10.1126/science.abd8755

    • Search Google Scholar
    • Export Citation
  • Rader, B. et al. (2020) Crowding and the shape of COVID-19 epidemics, Nature Medicine, 26(12): 182934. doi: 10.1038/s41591-020-1104-0

  • RKI (Robert Koch-Institut) (2020) Fallzahlen in Deutschland, https://npgeo-corona-npgeo-de.hub.arcgis.com/datasets/917fc37a709542548cc3be077a786c17_0/data.

  • Salvador, C., Berg, M.K., Yu, Q., San Martin, A. and Kitayama, S. (2020) Relational mobility predicts faster spread of COVID-19: a 39-country study, Psychological Science, 31(10): 123644. doi: 10.1177/0956797620958118

    • Search Google Scholar
    • Export Citation
  • Traunmüller, R. (2009) Religion und sozialintegration: eine empirische analyse der religiösen grundlagen sozialen kapitals (Abhandlungen), Berliner Journal für Soziologie, 19(3): 43568.

    • Search Google Scholar
    • Export Citation

Appendix A

Table A1:

Share of children receiving regular GPC and institutional childcare in percentages (columns 1–4), catholic population share in percentages (column 5), and foreign nationality population share in percentages (column 6) in 2017

StateShare of regular GPC (below age 15)Share of regular GPC > 7 hrs/weekShare institutional childcare, children 0–2Share institutional childcare, children 3–5Share Catholic populationShare foreign population
Baden-Wuerttemberg16.96.527.892.236.315.1
Bavaria16.67.226.589.753.812.7
Berlin12.44.643.890.48.817.6
Brandenburg14.15.555.592.33.34.4
Bremen13.45.825.982.810.917.4
Hamburg12.53.743.585.99.916.2
Hesse16.27.729.290.524.315.7
Mecklenburg-Vorpommern11.95.655.893.83.24.3
Lower Saxony14.56.329.290.317.59.0
North Rhine-Westphalia15.27.625.788.840.912.8
Rhineland-Palatinate16.87.630.093.744.210.6
Saarland16.98.727.791.261.910.7
Saxony16.16.350.693.73.64.6
Saxony-Anhalt16.68.257.091.73.44.7
Schleswig-Holstein14.26.031.389.36.07.7
Thuringia18.39.053.494.77.74.5
Rural county – sparsely populated16.57.537.691.527.76.2
Rural county18.58.734.391.131.27.4
Urban county17.28.028.290.836.911.5
Urban county (single metropolitan area)13.25.334.889.527.417.3
Total15.66.932.690.630.111.7

Notes: For columns 1–2, data from KiBS Survey, wave 6, 33,259 observations (weighted) (see Alt et al, 2020), lower number of observations due to missing information for weight calculation. For columns 3–6, data from BBSR (2020). Reference year for all data is 2017.

Table A2:

Share of children receiving regular GPC in percentages in 2017 (by age group)

Share of regular GPC
StateAge 0–2Age 0–2, > 7 hrs/weekAge 3–5Age 3–5, > 7 hrs/weekage 6–14Age 6–14, > 7 hrs/week
Baden-Wuerttemberg15.89.122.07.815.65.0
Bavaria15.99.523.010.214.55.4
Berlin11.63.915.86.011.44.3
Brandenburg11.94.420.38.912.74.8
Bremen14.17.813.55.113.05.1
Hamburg11.44.317.55.611.02.8
Hesse14.59.418.87.515.97.1
Mecklenburg-Vorpommern8.43.513.26.412.76.1
Lower Saxony16.69.217.87.412.74.9
North Rhine-Westphalia19.312.517.17.913.05.6
Rhineland-Palatinate16.18.322.411.315.16.2
Saarland17.711.123.510.914.47.2
Saxony12.94.821.27.815.56.1
Saxony-Anhalt14.87.019.79.416.28.1
Schleswig-Holstein15.77.719.67.811.94.7
Thuringia15.76.021.611.118.19.6
Rural county – sparsely populated17.010.120.59.715.36.3
Rural county20.712.525.011.316.06.8
Urban county17.310.622.39.215.66.8
Urban county (single metropolitan area)13.77.016.46.711.53.7
Total15.98.919.68.314.15.6

Notes: Data from KiBS Survey, wave 6, 33,259 observations (weighted) (see Alt et al, 2020), lower number of observations due to missing information for weight calculation.

Table A3:

SARS-CoV-2 infections and associated deaths by federal state according to RKI registers from 23 March 2020

StateInfections (total)Infections (per 100,000 inhabitants)Infections (age 60+)Infections (per 100,000, age 60+)
Baden-Wuerttemberg15,330138.54,114185.6
Bavaria18,251139.64723178.2
Berlin2,90179.645464.9
Brandenburg1,05542.026644.0
Bremen31245.78156.1
Hamburg2,578140.0473139.6
Hesse3,57657.181163.1
Mecklenburg-Vorpommern42526.49424.2
Lower Saxony4,76759.71,26572.7
North Rhine-Westphalia15,67487.43,777100.9
Rhineland-Palatinate3,06975.167176.4
Saarland94695.522496.6
Saxony2,28856.158555.5
Saxony-Anhalt77935.322038.3
Schleswig-Holstein1,27444.037757.0
Thuringia88641.322140.8
Total74,11189.018,356107.8

Notes: For registered SARS-CoV-2 infections, data are from RKI (2020). Own calculation of infections per 100,000 population by age group are based on BBSR (2020).

Table A4:

SARS-CoV-2 infections and associated deaths by federal state according to RKI registers from 30 September 2020

StateInfections (total)Infections (per 100,000 inhabitants)Infections (age 60+)
Baden-Wuerttemberg49,203444.5570.6
Bavaria67,761518.2615.4
Berlin14,326393.1314.1
Brandenburg4,251169.2181.0
Bremen2,385349.2272.3
Hamburg7,750420.9474.1
Hesse18,788299.8291.8
Mecklenburg-Vorpommern1,16872.672.4
Lower Saxony20,025250.9259.0
North Rhine-Westphalia69,283386.4383.1
Rhineland-Palatinate10,629260.2254.0
Saarland3,296332.8454.3
Saxony7,151175.4206.5
Saxony-Anhalt2,613118.3118.4
Schleswig-Holstein4,723163.0177.2
Thuringia4,056189.3258.9
Total287,408339.4383.3

Notes: For registered SARS-CoV-2 infections, data are from RKI (2020). Own calculation of infections per 100,000 population by age group are based on BBSR (2020).

Table A5:

Regressions explaining log registered infections per 100,000 (60+) (23 March 2020) at the county level

Log registered infections per 100,000 (60+)
Log days since first case1.428***0.892***0.885***0.914***0.889***
(0.169)(0.187)(0.184)(0.186)(0.174)
Share of regular GPC below age 15, > 7 hrs/week1.1891.1751.1271.2210.433
(1.061)(1.031)(1.009)(1.020)(0.969)
Log population/km2–0.0709–0.135**–0.0668–0.105
(0.0511)(0.0592)(0.0719)(0.0712)
Log median income1.792***1.339**1.484*1.161
(0.587)(0.628)(0.759)(0.709)
Eastern Germany–0.09090.0211–0.05290.345
(0.151)(0.151)(0.222)(0.271)
Urban county0.373***0.318**0.293**
(0.133)(0.138)(0.128)
Population share age 60+–1.509–1.4622.304
(2.632)(2.796)(2.807)
Population share under 187.6899.066*10.34**
(4.744)(5.243)(5.009)
Share institutional childcare (below age 3)0.0002280.00172
(0.00719)(0.00938)
Share institutional childcare (age 3–5)0.01830.0168
(0.0143)(0.0154)
Foreigner share–1.4480.705
(1.726)(1.635)
Catholic population share1.439***
(0.260)
Observations (counties)197197197197188

Notes: Minimum number of individual-level observations per county in KiBS survey = 30; regression estimates weighted based on population size. Standard errors in parentheses. *** p < 0.01; ** p < 0.05; * p < 0.1.

Table A6:

Regressions explaining log registered infections per 100,000 (60+) (30 September 2020) at the county level

Log registered infections per 100,000 (60+)
Log days since first case1.223***0.556***0.538***0.486***0.477***
(0.169)(0.181)(0.176)(0.176)(0.163)
Share of regular GPC, below age 151.597**1.932***1.851***1.609**0.570
(0.715)(0.664)(0.646)(0.651)(0.609)
Log population/km20.05670.00316–0.0670–0.126*
(0.0492)(0.0565)(0.0678)(0.0666)
Log median income1.132**0.679–0.0193–0.421
(0.565)(0.599)(0.717)(0.662)
Eastern Germany–0.280*–0.154–0.03460.338
(0.145)(0.143)(0.210)(0.255)
Urban county0.344***0.377***0.349***
(0.126)(0.130)(0.119)
Population share age 60+–1.316–0.6252.216
(2.510)(2.641)(2.616)
Population share under 1810.30**9.030*9.768**
(4.524)(4.952)(4.678)
Share institutional childcare (below age 3)–0.00710–0.00752
(0.00684)(0.00882)
Share institutional childcare (age 3–5)0.004660.00622
(0.0136)(0.0144)
Foreigner share3.339**5.649***
(1.629)(1.530)
Catholic population share1.158***
(0.240)
Observations (counties)197197197197188

Notes: Minimum number of individual-level observations per county in KiBS survey = 30; regression estimates weighted based on population size. Standard errors in parentheses. *** p < 0.01; ** p < 0.05; * p < 0.1.

Table A7:

Regressions explaining log registered infections per 100,000 inhabitants (60+) (30 September 2020) at the county level (Western Germany)

Log registered infections per 100,000 (60+)
Catholic population < 20%Catholic population ≥ 20%
Log days since first case–0.214–0.2570.771***0.714***
(0.463)(0.478)(0.182)(0.181)
Share of regular GPC, below age 150.7400.9170.5200.461
(2.083)(2.144)(0.691)(0.679)
Log population/km2–0.325*–0.362*0.01850.0300
(0.190)(0.208)(0.0845)(0.0831)
Log median income1.3951.525–0.161–0.0916
(1.803)(1.848)(0.764)(0.751)
Urban county0.642**0.687**0.03100.123
(0.287)(0.307)(0.160)(0.164)
Population share age 60+5.4154.518–5.201–2.935
(6.443)(6.798)(3.751)(3.841)
Population share under 185.1393.8772.2055.473
(12.76)(13.20)(5.148)(5.293)
Share institutional childcare (below age 3)0.01990.0182–0.0211*–0.0211*
(0.0229)(0.0235)(0.0107)(0.0108)
Share institutional childcare (age 3–5)–0.0257–0.01880.01760.0217
(0.0336)(0.0371)(0.0195)(0.0193)
Foreigner share6.2797.5020.06661.489
(4.862)(5.560)(1.969)(2.051)
Catholic population share–1.6510.800**
(3.475)(0.387)
Observations (counties)40409797

Notes: Only Western German counties with at least 30 individual observations in KiBS survey data are included, population–weighted coefficient estimation. *** p < 0.01, ** p < 0.05, * p < 0.1, standard errors in parentheses.

Appendix B

The exact wording of the question on GPC support in KiBS wave 6 is as follows: ‘In welchem Umfang wird Ihr Kind von den Großeltern betreut?’ (‘How often is your child looked after by his or her grandparents?’). Answers were as follows:

  • Regelmäßig, mit ___ Stunden in einer typischen Woche’ (‘Regularly, with __ hours in a typical week’)

  • Nach Bedarf’ (‘As required’)

  • Gar nicht’ (‘Not at all’)

  • View in gallery

    Illustration of causal mechanisms explaining infection rates at the county level

  • Adda, J. (2016) Economic activity and the spread of viral diseases: evidence from high frequency data, Quarterly Journal of Economics, 131(2): 891941. doi: 10.1093/qje/qjw005

    • Search Google Scholar
    • Export Citation
  • Albertini, M. and Kohli, M. (2013) The generational contract in the family: an analysis of transfer regimes in Europe, European Sociological Review, 29(4): 82840. doi: 10.1093/esr/jcs061

    • Search Google Scholar
    • Export Citation
  • Alipour, J.V., Falck, O. and Schüller, S. (2020) Homeoffice während der pandemie und die implikationen für eine Zeit nach der Krise, Ifo Schnelldienst, 73(7): 306.

    • Search Google Scholar
    • Export Citation
  • Alt, C., Anton, J., Gedon, B., Hubert, S., Hüsken, K., Lippert, K. and Schickle, V. (2020) DJI-Kinderbetreuungsreport 2019. Inanspruchnahme und Bedarf aus Elternperspektive im Bundesländervergleich, München: DJI.

    • Search Google Scholar
    • Export Citation
  • Althammer, J. (2013) Caritas in veritate: Katholische Soziallehre im Zeitalter der Globalisierung, Berlin: Duncker & Humblot.

  • Aparicio, A. and Grossbard, S. (2020) Intergenerational residence patterns and COVID-19 fatalities in the EU and the US, IZA Discussion Paper No. 13452.

    • Search Google Scholar
    • Export Citation
  • Arpino, B., Bordone, V. and Pasqualini, M. (2020) No clear association emerges between intergenerational relationships and COVID-19 fatality rates from macro-level analyses, Proceedings of the National Academy of Sciences (PNAS), 117(32): 19116121. doi: 10.1073/pnas.2008581117

    • Search Google Scholar
    • Export Citation
  • Aust, F., Burg, J.v.d., Hess, D. and Knecht, B. (2018) Methodenbericht – Kinderbetreuungsstudie 2017 und Zusatzuntersuchung ‚Pilotstudie: Räumliche Mobilität von Familien und Kindern in Zeiten der Digitalisierung, Bonn: Infas.

    • Search Google Scholar
    • Export Citation
  • Balbo, N., Billari, F.C. and Melgrano, A. (2020) The strength of family ties and COVID-19, Contexts, American Sociological Association, https://contexts.org/blog/structural-shocks-and-extreme-exposures/#balbo.

    • Search Google Scholar
    • Export Citation
  • Bayer, C. and Kuhn, M. (2020) Intergenerational ties and case fatality rates: a cross-country analysis, IZA Discussion Paper No. 13114.

  • BBSR (Bundesinstitut für Bau-, Stadt- und Raumforschung im Bundesamt für Bauwesen und Raumordnung [BBR]) (2020) Indikatoren und Karten zur Raum- und Stadtentwicklung (INKAR), Bonn: BBSR.

    • Search Google Scholar
    • Export Citation
  • Bouffanais, R. and Lim, S.S. (2020) Cities – try to predict superspreading hotspots for COVID-19, Nature, 583(7816): 3525. doi: 10.1038/d41586-020-02072-3

    • Search Google Scholar
    • Export Citation
  • Cohen, S. (2020) Psychosocial vulnerabilities to upper respiratory infectious illness: implications for susceptibility to coronavirus disease 2019 (COVID-19), Perspectives on Psychological Science, 16(1): 114.

    • Search Google Scholar
    • Export Citation
  • Dowd, J.B., Block, P., Rotondi, V. and Mills, M.C. (2020) Dangerous to claim ‘no clear association’ between intergenerational relationships and COVID-19, PNAS, 117(42): 259756. doi: 10.1073/pnas.2016831117

    • Search Google Scholar
    • Export Citation
  • Frerk, C. (2018a) Deutschland (2): Der evangelische Norden, https://fowid.de/meldung/deutschland-2-evangelischer-norden.

  • Frerk, C. (2018b) Deutschland (3): Der katholische Süden und Westen, https://fowid.de/meldung/deutschland-3-katholische-sueden-und-westen.

  • Frieden, T.R. and Lee, C.T. (2020) Identifying and interrupting superspreading events – implications for control of severe acute respiratory syndrome coronavirus 2, Emerging Infectious Diseases, 26(6): 105966. doi: 10.3201/eid2606.200495

    • Search Google Scholar
    • Export Citation
  • Gundlach, G. (1964) Die Ordnung der Menschlichen Gesellschaft (2 Bde), Köln: Bachem.

  • Hank, K. and Buber, I. (2009) Grandparents caring for their grandchildren: findings from the 2004 survey of health, ageing, and retirement in Europe, Journal of Family Issues, 30(1): 5373. doi: 10.1177/0192513X08322627

    • Search Google Scholar
    • Export Citation
  • Hippich, M. et al. (2020) Public health antibody screening indicates a six-fold higher SARS-CoV-2 exposure rate than reported cases in children, Med, 2(2): 115.

    • Search Google Scholar
    • Export Citation
  • Jappens, M. and van Bavel, J. (2012) Regional family norms and child care by grandparents in Europe, Demographic Research, 27(4): 84120. doi: 10.4054/DemRes.2012.27.4

    • Search Google Scholar
    • Export Citation
  • Kashnitsky, I. and Aburto, J.M. (2020) COVID-19 in unequally ageing European regions, World Development, 136: 105170. doi: 10.1016/j.worlddev.2020.105170

    • Search Google Scholar
    • Export Citation
  • Laxminarayan, R., Wahl, B., Dudala, S.R., Gopal, K., Neelima, S., Reddy, K.J., Radhakrishman, J. and Lewnard, J.A. (2020) Epidemiology and transmission dynamics of COVID-19 in two Indian states, Science, 370(6517): 6917. doi: 10.1126/science.abd7672

    • Search Google Scholar
    • Export Citation
  • Lee, E.C., Wada, N.I., Grabowski, M.K., Gurley, E.S. and Lessler, J. (2020) The engines of SARS-CoV-2 spread, Science, 370(6515): 4067. doi: 10.1126/science.abd8755

    • Search Google Scholar
    • Export Citation
  • Rader, B. et al. (2020) Crowding and the shape of COVID-19 epidemics, Nature Medicine, 26(12): 182934. doi: 10.1038/s41591-020-1104-0

  • RKI (Robert Koch-Institut) (2020) Fallzahlen in Deutschland, https://npgeo-corona-npgeo-de.hub.arcgis.com/datasets/917fc37a709542548cc3be077a786c17_0/data.

  • Salvador, C., Berg, M.K., Yu, Q., San Martin, A. and Kitayama, S. (2020) Relational mobility predicts faster spread of COVID-19: a 39-country study, Psychological Science, 31(10): 123644. doi: 10.1177/0956797620958118

    • Search Google Scholar
    • Export Citation
  • Traunmüller, R. (2009) Religion und sozialintegration: eine empirische analyse der religiösen grundlagen sozialen kapitals (Abhandlungen), Berliner Journal für Soziologie, 19(3): 43568.

    • Search Google Scholar
    • Export Citation
  • 1 German Youth Institute, , Germany

Content Metrics

May 2022 onwards Past Year Past 30 Days
Abstract Views 1 1 0
Full Text Views 2 3 3
PDF Downloads 3 3 3

Altmetrics

Dimensions