Waseda Institute for Advanced Study (WIAS)早稲田大学 高等研究所

News

Quantitative Comparison of COVID-19 Measures in Japan and the Rest of the World — Analysis of the Relationship between Social Issues and Politics/Economics to Clarify the Impact of COVID-19 Measures on Life in Society
ANNAKA Susumu, Assistant professor

ANNAKA Susumu, Assistant professor

Can the current system really reduce the number of COVID–19 deaths?

As of October 2020, one after another, countries around the world have launched policies and systems to combat the novel coronavirus infection (COVID-19). Among the daily reports issued by the mass media on related policies and expert opinions from around the world, some of those reports do not inspire confidence.

For example, one television news program argued that the family doctor system was one of the factors in the suppression of the fatality rate of COVID-19 in Germany. However, a similar system was in operation at the same time in Italy, which then had the highest number of deaths in the world. That raises the question as to whether Germany’s low case fatality rate may have been due to other factors.

In order to answer such questions accurately, I am conducting extensive statistical and quantitative analysis of rather a lot of data. Most importantly, rather than conducting medical analyses, I am analyzing social issues from a political economy perspective, studying the effect of policies and institutions on social problems.

Relationships that only become clear when we see the quantitative analysis

The case of the German family doctor system led to discussion based on the assumption that Germany’s system was effective, which in turn was based on the news analysis of only a part of the German system. In that light, the co-author and I began with an analysis of a body of data from 78 countries.

For the study, we used data from reliable sources such as the WHO database and data published by Johns Hopkins University; we conducted statistical analysis of the relationship between number of COVID-19 deaths and socioeconomic variables (policy, system, and social situation—Table 1 shows the case fatality rate (CFR), the ratio of deaths to positives, and Table 2 shows the result of deaths per 100,000 population. We conducted those two types of analysis, one for CFR and one for deaths per 100,000 population, in order to identify socioeconomic and demographic variables common to the two variables and to draw verifiable conclusions.

Variables in Tables 1 and 2 whose coefficients are marked with * or + bear some correlation with the number of deaths, and when the coefficient is positive (negative), an increase (decrease) in that variable leads to an increase in number of deaths. In both Tables 1 and 2, only two variables had the same coefficient sign and were marked with * or + : “hospital bed utilization ratio” and “population ratio over 65 years old.” From that analysis, we concluded that the low COVID-19 death toll was due to the socioeconomic and demographic factors “high number of hospital beds” and “low rate of population age 65 and over.”

This analysis was completed March 31, 2020. In the future, I aim to examine the recent situation as well as analyzing the impact of number of doctors and political system (degree of democracy).*

Table 1:

Relationship between socio-economic and demographic variables estimated by multiple regression analysis and CFR (logarithmic scale)

Coefficient Standard error
Hospital bed ratio (log) -0.4813* 0.1870
GDP (log) -0.3505* 0.1380
Population ratio of age 65 and over (log) 0.8136** 0.2472
Population density (log) 0.1525* 0.0761
Ratio of smokers 0.0029 0.0117
Number of days to present since confirmation of first infected person -0.0012 0.0070
Number of infected persons per 100,000 population. 0.0012 0.0018
Intercept -1.7383+ 1.0099
Adjusted R2 0.2017
Number of countries 78

+p,0.1;*p,0.05;**p,0.01;***p,0.001

Table 2:

Relationship between socio-economic demographic variables (estimated by multiple regression analysis) and number of deaths per 100,000 population (logarithmic scale)

Coefficient Standard error
Hospital bed ratio (log) -0.6565+ 0.2727
GDP (log) 0.7182** 0.2013
Population ratio of age 65 and over (log) 1.2342** 0.3605
Population density (log) 0.1339 0.1109
Ratio of smokers 0.0396* 0.0171
Number of days to present since confirmation of first infected person -0.0092 0.0101
Number of infected persons per 100,000 population. 0.0101*** 0.0027
Intercept -10.9169*** 1.4728
Adjusted R2 0.6474
Number of countries 78

+p,0.1;*p,0.05;**p,0.01;***p,0.001

In Table 1, the coefficient is negative and the result can be summarized as “CFR tends to decrease as GDP increases.” In Table 2, the result can be summarized as “Number of deaths per 100,000 population increases as GDP increases.” The analyses in Tables 1 and 2 are different, so it is not possible here to confirm the correlation between GDP and number of deaths.

(The above paragraph is quoted from International Comparative Analysis of COVID-19 Mortality Factors by Mogi and Annaka, WIAS Discussion Paper No. 2020-001 https://www.waseda.jp/inst/wias/assets/uploads/2020/05/001_Mogi-and-Annaka-2020.pdf )

We conducted another analysis using data from 151 countries to determine what kind of events will trigger the issuance of policy. Figure 1 shows what kind of COVID-19 situation triggered the implementation of non-medical government intervention policies (non-pharmaceutical interventions, hereinafter NPIs). NPIs include movement restrictions, requests for store closures, and penalties.

The data points in Fig. 1 show the relationship between number of deaths and number of positives (confirmed) each day before the interventions. The y-axis value is strength of effect of number of deaths and positives on NPIs. In Fig. 1, comparison of the two variables on the same day (e.g, -5 deaths and -5 confirmed cases) reveals that NPIs tend to be launched in response to number of positives rather than number of deaths, and were immediately put into practice (when the number of lagged days was small).

Figure 1: Relationship between Number of Deaths/Positives and Strength of NPIs
The analysis of data from Coronavirus Government Response Tracker (published by the University of Oxford at https://www.bsg.ox.ac.uk/research/research-projects/coronavirus-government-response-tracker) shows that policies and systems are launched in response to number of positives.
(Annaka, Susumu. 2020. COVID-19 and Government Responses. SocArXiv. July 23. Doi: 10.31235 /osf.io/xkswf. (https://osf.io/preprints/socarxiv/xkswf/)

Figure 2 also shows the results of an analysis of that daily data, i.e. how many days after the start of NPIs reduction of the number of deaths appears. The results show that even if NPIs are implemented, sufficient effects cannot be obtained until approximately 20 to 30 days later. In addition, the effect continues weakly for several weeks.

Figure 2: Effect of NPIs on Death Toll
It takes 20–30 days for NPIs to have an effect on number of deaths. On the other hand, number of deaths appears to increase immediately after the implementation of NPIs (the y-axis value is larger than zero). It is considered that this is because even if NPIs are implemented in response to increased number of positive tests, as shown in Fig. 1, it is inevitable that some of the positive-testing individuals will die.
Annaka, Susumu. 2020. “COVID-19 and Government Responses.” SocArXiv. July 23.
Doi: 10.31235/osf.io/xkswf. (https://osf.io/preprints/socarxiv/xkswf/)

Objective analysis of social issues by means of experimentation and cliometrics

Statistical analysis of COVID-19 is now my major research theme, but I am interested in various other themes and the use of various research methods.

For example, in one large scale survey study, I elucidated that the kind of tax system people prefer depends on their income, political party affiliation, and ideological beliefs.

I am also pursuing cliometrics using data such as records of testimony in war trials and Ministry of Interior statistics from the Meiji and Taisho eras. I have been unraveling anecdotal reports using data and historical records. I will continue to pursue this kind of study.

I decided to pursue that domain of research because I was interested in the relationship between policies and systems in response to issues such as North Korean poverty and the Lehman shock crisis. I intend to continue my research towards objective analysis of human life and social systems.

[Note] This article is scheduled to be published in the academic journal Public Choice.

Interview and composition: OISHI, Kaori
In cooperation with: Waseda University Graduate School of Political Science J-School

Page Top
WASEDA University

早稲田大学オフィシャルサイト(https://www.waseda.jp/inst/wias/)は、以下のWebブラウザでご覧いただくことを推奨いたします。

推奨環境以外でのご利用や、推奨環境であっても設定によっては、ご利用できない場合や正しく表示されない場合がございます。より快適にご利用いただくため、お使いのブラウザを最新版に更新してご覧ください。

このままご覧いただく方は、「このまま進む」ボタンをクリックし、次ページに進んでください。

このまま進む

対応ブラウザについて

閉じる