Identification of COVID-19 spread factors in Europe based on causal analysis of medical interventions and socio-economic data

Cover Page

Cite item


Since the appearance of COVID-19, a huge amount of data has been obtained to help understand how the virus evolved and spread. The analysis of such data can provide new insights which are needed to control the progress of the epidemic and provide decision-makers with the tools to take effective measures to contain the epidemic and minimize the social consequences. Analysing the impact of medical treatments and socioeconomic factors on coronavirus transmission has been given considerable attention. In this work, we apply panel autoregressive distributed lag modelling (ARDL) to European Union data to identify COVID-19 transmission factors in Europe. Our analysis showed that non-medicinal measures were successful in reducing mortality, while strict isolation virus testing policies and protection mechanisms for the elderly have had a positive effect in containing the epidemic. Results of Dumitrescu-Hurlin paired-cause tests show that a bidirectional causal relationship exists for all EU countries causal relationship between new deaths and pharmacological interventions factors and that, on the other hand, some socioeconomic factors cause new deaths when the reverse is not true.

Full Text

1. Introduction In January 2020, the SARS-CoV-2 coronavirus from 2019 made its way to Europe. As a result, the European Union and the majority of European nations had documented their first case. It should be observed, nevertheless, that the infection spread unevenly. At the end of April, there were more than three million confirmed cases of the severe acute respiratory syndrome coronavirus (COVID-19) worldwide (CSS, 2020), (SARS-CoV-2). The first human instance of the coronavirus was discovered in Wuhan, China, in late 2019 despite the fact that its origins are still unknown. One way the coronavirus is spread from person to person is through respiratory droplets created when infected people cough or sneeze in front of others [1]. Air travel is one of the factors contributing to the coronavirus outbreak in Europe. Late January or early February saw the confirmation of the first instances. Human contacts after the virus’s introduction to Europe helped it spread quickly. Social contact is crucial for the spread of all viruses, including COVID-19, according to research [2]. Human behavior is frequently viewed as a crucial safeguard for stopping the COVID-19 pandemic [3]. Globally, policymakers and health professionals are urging people to exercise social responsibility by limiting social interaction, adhering to stringent cleanliness and distancing guidelines, and being vaccinated. 1 Politicians are advising their constituents to weigh the social costs of their individual acts in terms of economics. In order to counteract COVID-19, official strategies heavily rely on this method of using social capital. The significance of social capital to controlling COVID-19 and preserving population health, however, is not well supported by systematic studies. According to what we know, this study is the first to rigorously analyze the dynamic link between social capital and health outcomes, as determined by COVID-19 instances and excess mortality. We systematically demonstrate that social capital has a causal and beneficial impact on pandemic-related health outcomes based on different analyses for seven European nations: Austria, Germany, Great Britain, Italy, the Netherlands, Sweden, and Switzerland. Personal hygiene habits and non-pharmaceutical interventions are the only ways to stop the spread of COVID-19 in the absence of vaccines and medications. The development of a broad framework for the causal analysis of COVID-19 in Europe is the goal of this research. As response variables, the number of new cases and fatalities attributable to COVID-19 are used. Potential causative variables include intervention factors and measures. 2. Related works Several studies have used various approaches and linked data from the WHO and other COVID-19 data sources to study the pandemic’s spread or serve as a guide for developing measures. Using the COVID-19 government response tracker data from the University of Oxford, employed Nonlinear Additive Noise Models for Bivariate Causal Discovery to determine the causative effect of a factor or an intervention measure on the number of new cases or an intervention measure. Reference [4] used data from the pandemic that affected 31 provinces and regions in China from January 20, 2020, to February 24, 2021, and the directed acyclic graph to demonstrate the causal link between influencing factor and daily cases. Using information from the official reports of the Robert Koch Institute, [5] studied the spread of the virus in Germany and the causative influence of restriction measures. In order to estimate the total causal effects based on directed acyclic graph analysis by negative binomial regression, collected data for 401 German districts between 15 February and 8 July 2020 from publicly accessible sources in Germany (e.g., the Robert Koch Institute, Germany’s National Meteorological Service, Google). The most commonly used statistical methods for analysing epidemiological factors of COVID-19 and evaluating intervention measures include correlation, regression, logistic regression and a dynamic model coupled with a linear model. Yet, if particular structures are considered, statistical methods like regression can only be regarded as instruments for causal analysis because they only allow a measure of causal dependence to be defined for these structures. On the basis of natural hypotheses, a procedure that is more effective than those now in use can be developed. Based on association analysis, this technique is known as dependency analysis. The statistical examination of the impacts of influencing factors and health interventions on the dissemination of COVID-19 has used association analysis as a reference. Yet, it is still challenging to comprehend the COVID-19 transmission pathway based on association analysis. The data were taken from the website used Pearson correlation analysis and multivariate linear regression to uncover economic and socio-political aspects that could fuel the coronavirus’s expansion. 3. Materials and methods 3.1. Data Description The analysis includes data for European economies from February 1st, 2020, through November 27th, 2022. Based on the statistics that are available, the era and the group of nations are chosen. The University of Oxford’s COVID-19 government response was where the information came from. The Government Response Index can be created using the data in this set, which also includes a stringency index, a containment and health index, and an economic support index (see table 1). Table 1 Definition of variables Variables Definition NEW_DEATHS News recorded deaths of COVID 19 STRINGENCY Stringency Index CONTAINMENT Containment Health Index ECONOMIC_SUP Economic Support Index VACCINATION Vaccination policy TESTING Availability of detection PROTECT_ELD Care policy for the elderly population The stringency index collects data on social segregation measures, coded from eight indicators: stay-at-home regulations, workplace closures, public event cancellations, gathering size restrictions, closures to public transportation, and travel restrictions both domestically and internationally. Three indices that stand for public awareness efforts, testing regulations, and contact tracing make up the containment and health index. The index stands for the government’s emergency health system policies, including the coronavirus testing program. The government’s income support program for citizens in times of crisis is reflected in the economic support index, which consists of two indicators: household anticipated debt alleviation and government income assistance. Each of these three metrics is expressed as a simple sum of the values for the underlying metrics, scaled to a range between 0 and 100. These indexes are provided for comparison and shouldn’t be used as a judgment on the suitability or efficacy of a nation’s approach. The WHO is the source of the daily total of new cases. The time frame for the study is from January 1, 2020, to December 4, 2022, and it includes 230 different nations. Table 2 displays a statistical breakdown of the key variables. The greatest value is 1623, the minimum value is 1918, and the average value is 42.27578, using the daily number of new deaths as an example. The number of new deaths is chosen as the explanatory variable since all efforts implemented by different governments around the world aim to prevent mortality, and reducing the number of cases will likely result in a decrease in deaths. So, the analysis will show us which measures not new instances as was noted in earlier literature really had an impact on pandemic related deaths. Table 2 Descriptive and Summary Statistics Variables Mean Standard Deviation Minimum Maximum NEW_DEATHS 42.27578 104.4784 -1918.000 1623.000 STRINGENCY 43.11964 23.07557 0 96.30000 CONTAINMENT 49.82720 17.54525 0 90.00000 ECONOMIC_SUP 57.41835 34.87956 0 100.0000 VACCINATION 2.998873 2.247651 0 5.000000 TESTING 2.355943 0.799513 0 3.000000 PROTECT_ELD 1.588960 1.006744 0 3.000000 3.2. Methodology In our empirical research, we examined how health interventions and socioeconomic observational data contributed to the global spread of COVID19. Using this method, we may assess how health measures have affected the spread of COVID-19. In order to accomplish our goal, we used in this study a linear function that incorporates socioeconomic observational data and health treatments as an extra variable to control factors that are equivalent to COVID-19. As suggested by Pesaran and Shin, the equation is calculated using a time series autoregressive distributed lag model (ARDL). The advantage of the ARDL framework is that it can differentiate between short- and long-term impacts, which enhances earlier material. We may also predict a consistent short-term cross-sectional influence (short term coefficient of nations) due to our extensive sample size. Due to its distinction between short- and long-term impacts, the ARDL methodology aids in addressing the shortcomings of earlier work. Using both time and cross-sectional dimensions increases the overall number of data and their variability in our panel estimation. A panel estimation also reduces the noise that results from a single time-series estimation, leading to more trustworthy inference. 3.2.1. Panel unit root tests The determination of the order of integration of variables serves as the foundation for estimating any econometric model. It is required to verify that the variables in the regression are either integrated of order zero

About the authors

Kouame A. Brou

Peoples’ Friendship University of Russia (RUDN University)

Author for correspondence.
ORCID iD: 0000-0003-1996-577X

PhD student of Information Technology Department

6, Miklukho-Maklaya str., Moscow, 117198, Russian Federation


  1. S. L. Priyadarsini and M. Suresh, “Factors influencing the epidemiological characteristics of pandemic COVID 19: a TISM approach,” International Journal of Healthcare Management, vol. 13, no. 2, pp. 89- 98, 2020. doi: 10.1080/20479700.2020.1755804.
  2. A. Farseev, Y.-Y. Chu-Farseeva, Q. Yang, and D. B. Loo, “Understanding economic and health factors impacting the spread of covid-19 disease,” medRxiv, 2020. doi: 10.1101/2020.04.10.20058222. eprint:
  3. K. Tantrakarnapa, B. Bhopdhornangkul, and K. Nakhaapakorn, “Influencing factors of COVID-19 spreading: a case study of Thailand,” Journal of Public Health, vol. 30, pp. 621-627, 2020. doi: 10.1007/s10389-02001329-5.
  4. W.-X. Tang, H. Li, M. Hai, and Y. Zhang, “Causal Analysis of Impact Factors of COVID-19 in China,” Procedia Computer Science, vol. 199, no. 10229, pp. 1483-1489, 2022. doi: 10.1016/j.procs.2022.01.189.
  5. R. Chaudhry, G. Dranitsaris, T. Mubashir, J. Bartoszko, and S. Riazi, “A country level analysis measuring the impact of government actions, country preparedness and socioeconomic factors on COVID-19 mortality and related health outcomes,” eClinical Medicine, vol. 25, p. 100464, 2020. doi: 10.1016/j.eclinm.2020.100464.
  6. A. Levin, C.-F. Lin, and C.-S. J. Chu, “Unit root tests in panel data: asymptotic and finite-sample properties.,” Journal of Econometrics, vol. 108 (1), pp. 1-24, 2002. doi: 10.1016/S0304-4076(01)00098-7.
  7. M. H. Pesaran, T. Schuermann, and S. M. Weiner, “Modeling regional interdependencies using a global error-correcting macroeconometric model,” Journal of Business & Economic Statistics, vol. 22, pp. 129-162, 2004. doi: 10.1198/073500104000000019.
  8. M. Pesaran, Y. Shin, and R. J. Smith, “Bound Testing Approaches to the Analysis of Level Relationship.,” Journal of Applied Econometrics. Special Issue: In Memory of John Denis Sargan 1924-1996: Studies in Empirical Macroeconometrics, vol. 16, no. 3, pp. 289-326, 2001. doi: 10.1002/JAE.616.
  9. M. P. Clements and D. F. Hendry, “Forecasting in cointegrated systems,” Applied Econometrics, vol. 10, pp. 127-146, 1995. doi: 10.1002/jae.3950100204.
  10. S. Johansen, “Statistical analysis of cointegration vectors,” Journal of Economic Dynamics and Control, vol. 12, no. 2-3, pp. 231-254, 1988. doi: 10.1016/0165-1889(88)90041-3.
  11. R. Engle and C. Granger, “ Cointegration and Error Correction: Representation, Estimation and Testing.,” Econometrica, vol. 55, no. 2, pp. 251-276, 1987. doi: 10.2307/1913236.
  12. E.-I. Dumitrescu and C. Hurlin, “Testing for Granger non-causality in heterogeneous panels,” Economic Modelling, vol. 29, pp. 1450-1460, 2012. doi: 10.1016/j.econmod.2012.02.014.

Copyright (c) 2023 Brou K.A.

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

This website uses cookies

You consent to our cookies if you continue to use our website.

About Cookies