A new policy brief by Shoresh Institution for Socioeconomic Research vice president, Prof. Ayal Kimhi, “Socioeconomic aspects of Coronavirus infection rates across Israeli municipalities,” identifies the impact of several major municipal attributes on the infection rate, while controlling for the effects of other attributes.
Utilizing a multiple regression analysis, Prof. Kimhi examined a number of municipal characteristics that could potentially explain differences in infection rates across Israeli towns. He identified three municipal attributes with statistically significant (and robust to model specification) effects on the infection rate: the percent of population in the 75+ age group; population density per square kilometer of residential area; and the percent of the population living in religious boarding schools. Other attributes that were examined, but were not found to affect the infection rate significantly, include the municipality’s socioeconomic ranking, the share of population in the 35-55 age group with an academic degree, the average wage of salaried employees, and the municipality’s peripheriality cluster. Exclusion of these variables from the regression did not change the results in a meaningful manner.
Infection rates were found to be negatively correlated with the percent of population in the 75+ age group, perhaps because people in that age group followed the social isolation orders more strictly. However, this relationship was not particularly large. On the other hand, the other two variables had substantial quantitative effects on municipal infection rates, and their statistical significance was robust to various specifications of the regression model.
Figure 1 highlights the strong link between infection rates and the fraction of population living in religious boarding schools:
Figure 2 displays the strong relationship between infection rates and population density.
While the positive links are clear in both figures, it is also evident is that these relationships are stronger at the extreme values of each of the explanatory characteristics.
To highlight the contribution of these two municipal attributes to the extent of coronavirus infections, the Shoresh study highlighted two groups of Israeli municipalities at both ends of the infection spectrum. 1.3 million people live in municipalities with infection rates below 60 per population of 100,000, and 1.6 million people live in municipalities with infection rates above 300 per population of 100,000. The infection rate in municipalities with the lowest infection rates averaged 31 persons per population of 100,000 while in municipalities with the highest infection rates, the infection rate averaged 560 persons per population of 100,000.
Prof. Kimhi then simulated how an increase in the share of population living in religious boarding schools (from the low share in the low infection towns, to the high share in the high infection towns) would have contributed to higher coronavirus infection rates in the low infection towns. Specifically, infection rates in the low infection towns would have increased by 232 persons per population of 100,000.
The Shoresh study then conducted a similar simulation on the impact of population density within towns. In municipalities with the lowest infection rates, population density is about 9,300 persons per square kilometer. At the other end of the spectrum, municipalities with the highest infection rates possess population density over two and a half times higher – about 25,000 persons per square kilometer. An increase in population density in the towns with the lowest infection rates to the population density in the towns with the highest infection rates would have increased their infection rates by 205 persons per population of 100,000.
As shown in Figure 3, these two attributes explain nearly the entire gap in infection rates between the two groups of municipalities.
Prof. Ayal Kimhi surmises that the strong relationship between infection rates and the fraction of population living in religious boarding schools could be due to lack of compliance with the social distancing regulations – either because of simple misconstruing of these regulations, a basic attitude of non-compliance with government authorities, or to irresponsible conduct by the heads of those institutions. Prof. Kimhi: “It should be noted that no statistical relationship was found between infection rates and the percentage of population living in non-religious boarding schools, in medical institutions, in assisted living accommodations, and in other institutions. As for population density, it is not necessary to explain how it’s related with infection rates.”
Policy ramifications from the Shoresh study suggest a middle ground between the policy of overall shutdown that was imposed in Israel, with its enormous economic damage, and a policy of a selective shutdown, that needs to be based on a wide and efficient testing system that does not currently exist. Specifically, Prof. Kimhi states that it may be possible to use a selective shutdown policy in certain municipalities with certain attributes in which the risk of being infected and spreading the virus is particularly high.