Covid-19 in Israel: socio-demographic characteristics of first wave morbidity in Jewish and Arab communities

Israel’s first case of Covid-19 was diagnosed on February 26, 2020. On March 14, 2020, 200 cases were confirmed and additional 40–60 cases were being diagnosed daily. Two days later, on March 16, a general lockdown was declared. Mask wearing in public became obligatory on April 12th, but the requirement was hardly enforced for long weeks thereafter. Professionals offered polarized forecasts, ranging from a total of 10 deaths throughout the epidemic, to ten thousand deaths, accompanied by fatal shortage of ICU facilities, ventilators and medical staff. The first wave climaxed in early April with nearly 10,000 active cases and 500–700 additional confirmed cases a day. Since then, the spread started to slow down. Starting April 19, lockdown restrictions were gradually relaxed. As of early June 2020, Israel had roughly 2000 active cases, less than 300 deaths and a fatality rate of 1.67%, substantially lower than that of most West European countries and well below the world average of 6.15%.¹ At the time of this writing, a second wave of Covid-19 is plaguing Israel. The present study explores morbidity rates in Israel on a residential community basis, aiming to identify associations with main sociodemographic variables, as they evolved under the lockdown practice.

Scholars have suggested numerous variables that may affect the spread and scope of the pandemic, including medical resources, health coverage, life expectancy, employment rates, income inequality, levels of social connectedness and education, alongside religious affiliation, household size, population age composition, population density and ecological footprint [1]. Of this wide array, the present study focuses on four variables that have been repeatedly suggested in Israel’s public discourse, as influencing local morbidity rates: population density, socioeconomic status (SES), percentage of elderly population and minority status. We supplement these four variables by zooming in on one component of the SES index, that of household internet access, which was pointed out as associated with morbidity in Israel, due to inter-sector differences in access. We start with descriptive statistics of these variables. To the best of our knowledge, despite repeated references in the media and public discourse, no systematic analysis of these variables has been conducted to date. A subsequent multivariate analysis seeks to explain the associations between these variables and morbidity rates.

Socioeconomic status has often been mentioned in conjunction with rates of Covid-19 morbidity. In the UK, for instance, the disease appears to have spread along class lines. In the early stages of the outbreak, regions with higher SES were over-represented among the sick. Explanations attributed the excess to more frequent traveling and social contacts, alongside better access to testing [2]. Later on, however, infection rates started to rise in densely populated areas, especially those suffering higher levels of deprivation, affecting primarily people from lower socioeconomic urban areas, many of whom also belonged to ethnic minorities [3]. A similar picture emerged in New York City, where low-income neighborhoods were disproportionately hit by the pandemic [4] The contribution of SES to morbidity rates is thought to operate through various channels of disadvantage, such as posing hurdles to social distancing due to crowded housing, jobs that require physical presence and reliance on public transport for job attendance.²

Population density is another factor that has long been connected to epidemics. In the present pandemic, high morbidity was documented in dense European and American metropolises like London [3], Paris [5], Milan (https://​www.​citypopulation.​de/​en/​italy/​covid/​) and New York [6]. All these cities stood out in comparison to their less heavily populated surroundings, with substantially elevated rates of morbidity. Population density is, however, not the sole determinant of morbidity rates, as powerfully demonstrated by the low incidence of disease in Asian megacities like Hong Kong, Singapore, Seoul and Tokyo.

Age composition has been singled out in the present pandemic, as associated with morbidity rates [7]. High mortality was attributed to the presence of numerous elderly people, defining such age structure as creating vulnerability to Covid-19 and especially, to its complications. Social settings in which multi-generation households are prevalent have also been mentioned as more susceptible to contagion and subsequent death, primarily among older people [8].

Minority status is yet another factor suggested as associated with increased morbidity. In England, for instance, a high proportion of black and minority residents was found to be the strongest predictor of high Covid-19 death rate and people of ethnic minority descent were over-represented among Covid-19 victims [9]. Evidence in the same direction came from the U.S., where black and indigenous people sustained twice or more Covid-19 deaths per capita as compared to white Americans [10].

In the following, we explore the extent to which these factors – SES, population density, size of elderly population and minority status – were associated with morbidity rates in Israel’s residential communities.

This is a cross-sectional study, investigating all of Israel’s residential communities with more than 5000 people (i.e., full population rather than sample). It focuses on Israel’s cumulative Covid-19 morbidity figures since the beginning of national registration in February 23, 2020, up until June 2, 2020, when the first wave of the epidemic was considered to have been contained.

Looking at Israel’s morbidity rates by SES (Fig. 1), we observe a general trend, suggesting that a rise in SES is associated with a decrease in morbidity rates, i.e., higher morbidity in lower ranking communities and lower morbidity in wealthier ones.

Notably, the SES levels 2 and 3 sustained the highest morbidity rates in Israel. These categories include numerous Jewish orthodox communities (e.g. Bnei Brak, Elad, Kiryat Yearim, Emanuel, Rechasim in rank 2 and Beit Shemesh, and Zefat in rank 3), which have indeed suffered the highest morbidity in the country. SES level 1 is also populated by orthodox communities. However, it also includes numerous peripheral Arab Bedouin communities. These communities were relatively less affected than communities in the two subsequent levels, as illustrated in Fig. 1. SES category 4 does not include as many orthodox communities, but rather traditional peripheral Jewish as well as Arab settlements. These communities have apparently been relatively less affected than their poorer and wealthier counterparts, possibly due to their peripheral location and distance from major sites of contagion.

When assessed by the communities’ population density (Fig. 2), the morbidity distribution appears to follow an even clearer trend, generally increasing with the population density: the higher the density, the higher the morbidity. At the upper right hand corner, the city of Bnei Brak, with the highest population density in Israel – roughly 50% denser than the subsequent town (See Table 1) – sustained the highest morbidity rate in Israel during the epidemic’s first wave: 14.72 per 1000 persons. At the upper left corner, the small Arab community of Deer El Assad is an outlier, where morbidity soared for a relatively short period of time in April, but was contained within a few days.

Nonetheless, as suggested by Table 1, communities differed greatly. For example, though Elad’s population density is roughly 3/4 that of Givataim, morbidity rate in the former was 10 times that of the latter. Visibly, orthodox cities suffered much greater morbidity than denser non-orthodox communities. Morbidity rates thus increased with population density. Yet, it was clearly not a sole determining factor.

The third factor is the size of elderly population as measured by the proportion of people aged 65+ in the community. The distribution (Fig. 3) depicted an inverse relation between the variables, i.e., the higher the percentage of elderly persons in a community, the lower the morbidity. The two outliers are once again, the communities of Bnei Brak and Deer El Assad, which stood out in the density distribution above (Fig. 2).

The descrpetive statisitics thus suggest 1. Inverse association between a community’s SES and its morbidity rate; 2. Positive association between population density and morbidity; 3. Inverse relation between percentage of elderly population and morbidity.

The last result, which goes somewhat against findings reported from other settings, requires some more attention. We start with the city of Bnei Brak, that had the highest morbidity rate in Israel (14.7 per 1000) but a mere 8.9% of elderly residents. Jerusalem and Haifa, in contrast, with evidently larger proportions of elderly residents (12.3 and 27% respectively), had strikingly lower morbidity rates: 4.16 and 0.61 per 1000 respectively. A more systematic look into the communities with the highest morbidity rates (Table 2), shows that the proportions of elderly population in these communities were all below the average for the Jewish sector (13%) and even below the national average (11%), thereby illustrating the inverse association:

Notably, all four cities are orthodox Jewish communities.

Exploring the opposite direction, reveals similar findings. As shown in Table 3, the morbidity rates in Israel’s ‘oldest communities’, namely those with the largest proportions of people older than 65 (19-21%), were consistently below (0.32–1.51) the national average of 1.64 per 1000 at the time (June 2, 2020). None of these communities is orthodox.

At a glance, the low proportion of elderly people in Israel might be taken as a protective factor that helped moderate morbidity rates in the country. However, the findings suggest the opposite, as illustrated in Tables 2 and 3. The key factor explaining this unusual distribution is the exceptional fertility rates in Israel’s orthodox Jewish communities that stand on 7.1 children per woman, as opposed to roughly 3 children per woman in other Jewish and Arab communities. This difference drastically reduces the percentage of the orthodox communities’ elderly cohorts. At the same time, these very large families are mostly poor, often crowding in small apartments, and as such, are prone to contagion.

In order to gain a closer insight into the association of each variable with morbidity rates, we conducted a linear regression, wherein the dependent variable is the morbidity rate. The following results were obtained (Table 4):

The regression model thus predicts 38.1% (R² = 0.381) of the inter-community morbidity rate variance in Israel at the community level, by SES, density, elderly population and minority status. As shown, SES turned out to be statistically insignificant, but population density, size of elderly population and minority status emerged as significantly associated with morbidity rates. The association of each variable with COVID-19 communities’ morbidity rates is presented in Fig. 4, which shows the respective regression lines.

The nearly horizontal SES line visualizes the lack of statistically significant association between this variable and the rate of morbidity in Israel. Population density, in contradistinction, emerged as significant, affirming the public discourse that linked crowdedness with increased morbidity. Indeed, population density emerged as the strongest association with communities’ morbidity rates (β = 0.439). A 100 person increase per km², raises morbidity by 0.024 patients per thousand (b = 0.00024), namely, 2.4 additional patients per 100,000 persons. As for the percentage of a community’s elderly residents, as suggested by the descriptive statistics, the regression shows a significant inverse association between this variable and communities’ morbidity rates, i.e., the smaller the elderly population, the higher the local rate of morbidity (b = − 9.16). In terms of minority status, despite their lower SES and minority status, Arab communities sustained lower morbidity rates, i.e., Jewish communities suffered 1.68 higher morbidity rates compared to Arab communities (b = 1.68, β = 0.437). Moreover, the Jewish-Arab gap in morbidity proved persistent. Figures 5 and 6 present the marginal association of population density and percent of elderly population – the two variables that were found to be statistically significant – with morbidity rates in the Jewish and Arab populations.

The calculation is based on the regression model, assuming that all other variables are held constant. The figures demonstrate that the higher the population density, the higher the community’s morbidity rates and the larger the elderly population, the lower the morbidity rate. Equally important: at any given value of population density or size of elderly population, the gap in morbidity rate between the Jewish and Arab cities persists, with morbidity rates consistently lower in Arab communities.

In the light of the emergent gap in morbidity between Arab and Jewish communities, we divided our calculation and produced the average values of the scrutinized sociodemographic variables alongside Covid-19 morbidity rates in Jewish vs. Arab communities. Table 5 presents the averge values in each sector.

The table reveals a clear Jewish-Arab difference: with communities twice or more richer, denser and older, Israel’s Jewish cities sustained 2.6 times the morbidity of their Arab counterparts. Whereas these figures accord with international findings regarding population density and proportion of elderly residents, the higher morbidity in richer cities is exceptional. In other words, Israel’s Arab communities, despite being much poorer and despite belonging to an underprivileged minority, sustained less than half the morbidity of their wealthier Jewish neighbors. Therefore, we examined how these variables are associated with morbidity in Jewish vs. Arab communities. The linear regression analysis presented in Table 6, reveals further details and differences:

The analysis shows that when considered on its own, SES (Model 1) becomes significant in the Jewish communities though not in the Arab ones. Thus, in Israel’s Jewish communities SES was associated with morbidity in a manner similar to that observed elsewhere, i.e., poorer communities were more vulnerable to the virus than wealthier ones. More concretely, a rise in one SES rank in a Jewish community was associated with a reduction of 0.373 sick person per 1000. In the Arab sector, no such association was found. However, once the other variables are added (Model 2), the association of SES with morbidity once again becomes statistically insignificant also in the Jewish sector. Population density (Model 2), however, proved significant and equally influential in both sectors.

The third variable, proportion of elderly population, also emerged as statistically insignificant in the Arab communities but significant and inversely associated with morbidity in Jewish communities, i.e., a smaller elderly population was associated with higher morbidity. Another substantial inter-sectorial difference related to the explained variance. While explaining nearly 38% of the variance in morbidity rates in Jewish communities, Model 2 accounts for less than 11% in Arab settlements. The model thus reaffirmed the lay perception that population density was positively associated with morbidity rates. However, in contrast to international findings, SES was statistically insignificant to morbidity rates in the Arab communities and even in the Jewish sector, the association was weak.

Finally, some Israeli commentators attributed the high morbidity rates in some ultra-orthodox communities to the scarcity of home and mobile Internet access in these communities, that wish to maintain but minimal contact with their non-orthodox surrounding. (‘Kosher mobile phones’ have only pre-designated internet access, e.g., to banks, HMO etc.) This virtual detachment was assumed to underlie the communities’ lack of awareness and belated application of the protective measures and lockdown guidelines. We therefore zoomed in on this specific component of the SES index and conducted a specific linear regression.

The general regression revealed no association with internet access. Sector specific analyses (Table 7) found that home Internet access followed the SES pattern, namely, significant association with morbidity only in Jewish communities and only when considered on its own.

However, the latter association was substantial: a 10% increase in home internet access in Jewish communities was associated with a morbidity decline of 0.3 person per 1000. The resemblance to SES patterns is reasonable, given the fact that internet access is a component of the SES index.

Figure 7 further supports Table 7 findings, showing that as internet access rises, the communities’ morbidity rates decrease.

We now turn to discuss the observed trends and particularities and propose some lines of explanation.

The current analysis revealed increased morbidity rates in dense, low SES Jewish communities with small elderly populations and low access to household Internet. This description befits first and foremost Jewish Orthodox communities. As mentioned, these communities have indeed suffered the highest morbidity rates in Israel. More generally, population density was the only variable that emerged as statistically significant to morbidity in the entire Israeli population. In Arab communities, the lower density - 2037 vs. 3853 persons per km² – might have contributed to the lower morbidity. The association of SES with communities’ morbidity solely in Jewish cities probably reflects the relatively narrow range of SES distribution of Israel’s Arab communities, which are largely clustered at the lower end of the SES scale. However, in the Jewish sector, too, SES was significant only when analyzed on its own and then, too, the association was moderate: massive differences in morbidity were observed among communities with similar SES⁶ and similar morbidity rates were recorded in communities far apart in their SES ranking.⁷

The association between morbidity and the size of a community’s elderly population is probably the most peculiar finding of the present study. Israel’s population is much younger than that of other industrialized countries. The main reason for the gap is the local fertility rate of 3.11 children per woman, which is nearly twice the OECD average of 1.7.⁸ In 2015, people aged 65+ thus comprised 11% of Israel’s population, representing 13% of the Jewish population and a mere 4% of Israeli Arabs. For the sake of comparison, in Japan and Italy the elderly comprise 23 and 28% of the population, respectively.⁹ The OECD average is 17.2%. Having said that, we need to bear in mind that older people do not necessarily contract Covid-19 more often, but are rather at higher risk for complications. As such, the inverse association may not be as paradoxical as it may initially appear to be. In fact, the low proportion of elderly population might have contributed to Israel’s low mortality rate. Ibe. BeyondBeyond this particular difference, the analysis revealed that when density level, SES or size of elderly population are held constant as average, the prediction of morbidity rate in the Jewish communities is 1.68 times higher than in Israel’s Arab communities. This lower morbidity rate among the country’s poor minority population, as compared to the wealthier majority group, counters findings from other countries. Moreover, multi-generation households, another variable associated with higher morbidity, as mentioned, are more prevalent in Israel’s Arab communities than in Jewish ones [19]. How can we explain these surprising findings? What local particularities can account for this difference? We propose the following lines of explanation:

In contradistinction, Israel’s Arab communities, though belonging to an underprivileged, poorer minority (e.g., [22, 23]) and often living in multi-generation households, had several protective factors:

As of May 20, 2020, there were a total of 1040 confirmed Covid-19 cases and five deaths in Arab communities in Israel, since the outbreak of the epidemic [25].

To sum up, Arab communities in Israel were relatively less affected by the Covid-19 epidemic than Jewish communities, despite their relative poverty, minority status and prevalent multi-generational households. Lower population density and smaller families, alongside highly efficient community self-organization and collaboration with state bodies resulted in exceptionally low morbidity and mortality rates. In contradistinction, high morbidity was observed in Israel’s poor, crowded Orthodox communities, especially ones with very small elderly populations.

The observed trends and differences may possibly offer an insight about potential state measures and the protection and inclusion that they have offered to marginalized communities in the context of pandemics. A comparison with morbidity patterns during Israel’s second wave of Covid-19 morbidity may reveal how different containment measures have affected the health and lives of various subpopulations across the social ladder.