key: cord-0715710-uwix8ftr authors: civcir, i. title: Evaluation of Turkish social distancing measures on the spread of COVID-19 date: 2020-05-04 journal: nan DOI: 10.1101/2020.04.28.20083550 sha: e3e4b951ffebb16f34182e3d8951750e2c4e4ce3 doc_id: 715710 cord_uid: uwix8ftr The coronavirus disease (COVID-19) affecting across the globe. The government of different countries has adopted various policies to contain this epidemic and the most common were social distancing and lockdown. We use a simple log-linear model with intercept and trend break to evaluate whether the measures are effective preventing/slowing down the spread of the disease in Turkey. We estimate the model parameters from the Johns Hopkins University (2020) epidemic data between 15th March and 16th April 2020. Our analysis revealed that the measures can slow down the outbreak. We can reduce the epidemic size and prolong the time to arrive at the epidemic peak by seriously following the measures suggested by the authorities. postponed. On 20 March, all kinds of cultural and scientific activities/meetings were suspended. Free public transportation for elderly people was temporarily suspended in several metropolitan cities. On 21 March, the activities of barbershops, hairdressers, and beauty salons were stopped. Flights to in a total of 68 countries had stopped. A total curfew announced for those who are over the age of 65 or chronically ill. Restaurants, dining places, and patisseries were to be closed to the public for dining in and were only allowed to offer home delivery and take-away. On 27 March, all international flights were terminated, and intercity travel was subject to approval by the state governors. Furthermore, historical sites and picnic areas are closed on weekends. On 3 April 2020, the government issued a 15-day entry ban to 30 metropolitan cities as well as Zonguldak, which is extended later. Furthermore, the curfew was extended to people younger than 20 years old. Using masks in public places became mandatory ( https://en.wikipedia.org/wiki/2020_coronavirus_pandemic_in_Turkey). This article aims to determine the effectiveness of government measures to prevent or slow down the spread of Covid-19 in Turkey. We would anticipate the consequences of these measures to be visible as of 21 March 2020 and beyond. As expected, if these policy measures work to prevent or slow down the spread of Covid-19, the growth rate of the number of confirmed cases should display a slowdown starting from around 21 March. The structure of the paper is as follows. Section 2 describes the data for confirmed cases of in Turkey. In Section 3, a simple model set up to determine the average growth rate of log cumulative confirmed cases, and this model is tested for intercept and trend breaks. In Section 4, the estimation results are presented and discussed. Section 4 concludes. To understand the effects of the government measures on the spread of Covid-19 in Turkey, we investigate the development of the number of confirmed cases of Covid-19 in Turkey between 15 March and 26 April 2020. There are several sources of data. Official data can be obtained from the Ministry of Health (2020). Other main alternative sources are the World Health Organisation, the European Centre for Disease Prevention and Control and the Johns Hopkins University (2020). We used the Johns Hopkins University (2020) data, because it links data from the Health Ministry, the World Health Organization. There are slight outliers at the beginning of the sample, where the case number is low but the variation is relatively high, therefore, we started our investigation after 10 or more cases reported which start from 15 March. Between 15-20 March, the growth rate of cumulative confirmed cases is about 123%, which implies that a doubling of confirmed cases every 0.57 days. Between 21-26 March, the growth rate of the confirmed cases drops to 33%, implying that the doubling of confirmed cases every 2.13 days. Between 27 March and 1 April, the average growth rate dropped to 28%, and the doubling of confirmed cases improved to every 2.47 days. Between 2-11 April, the growth rate dropped to 13 percent, 12-19 April to 8 percent, and 20-26 April to 5 percent and the doubling of the confirmed cases improved to 14.5 days, which is in line with the lagged impact of the policy measures taken by the Turkish authorities. Officials and the general public knew that the virus was spreading around the world, they knew that they would face this problem, but they did not know when. After the first case was announced on 11 March, and with the initial rapid growth of the confirmed cases, the authorities implemented several policy measures to contain the rapid spread of Covid-19. The establishment of an independent scientific board consisting of faculty members of the medical schools and making public announcements after these committee meetings made the public follow the board's recommendations closely. The general public started to follow the developments closely from the media and this increased public awareness of Covid-19. Since there is a delay of approximately seven to eight days between virus infection and its reflection in statistics, the implemented policies cannot be expected to slow down the spread of Covid-19 immediately. The incubation period of the coronavirus is between 2 and 10 days, the mean estimate is 5 days (Lauer et al.; 2020; Linton et al.; 2020; WHO; 2020) . Therefore, we do not expect to see the impact of the implemented policies and thus changing individual behavior on the Covid-19 confirmed cases until March 21. When we visually examine the log confirmed cases in Figure 1 , we can see that the spread of Covid-19 has slowed down from March 21onwards. Nonetheless, Covid-19 data should be analyzed by an appropriate statistical method to understand whether the decrease in the number of case's growth rate is decreasing randomly or systematically. To investigate the impact of the measures taken by the authorities, we will investigate whether there are an intercept and trend breaks in Covid-19 data. As can be seen from Figure 1 , the log confirmed cases increase quite linearly in t with level shifts and trend breaks, therefore, we expect specification (1) to capture the developments in the confirmed case time series well. More complex models may be required to accurately determine the drivers of logapproved cases such as weather and seasonal effects. However, these factors are currently unlikely to be correlated with a linear trend, we expect the estimates of model (1) to hardly be affected. Additionally, the time series on confirmed cases is relatively short, impeding more complex modeling of the data. We included the lagged values of the endogenous variable, but they turned out to be insignificant. To get the growth rate of log confirmed cases, we initially estimated a model with intercept and trend by ordinary least squares, thus setting . For the simple model without intercept and trend breaks, our estimation results revealed that the intercept is ߙ ො ൌ All rights reserved. No reuse allowed without permission. was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (2020), we tested for a structural break in intercept and trend. However, we allowed for more than one unknown trend and intercept break. To determine the breaking point, we employed multiple unknown breakpoint tests developed by Bai and Perron (1998 , 2003a , 2003b and Yao (1988) information criterion based multiple breakpoint tests. Bai and Perron (1998 , 2003a , 2003b test is robust to serial correlation and heteroscedasticity. We adopted a general to specific methodology to determine the exact number of the break, i.e. we dropped the insignificant intercept and trend break variables from the selected model. All rights reserved. No reuse allowed without permission. was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which this version posted May 4, 2020 . . https://doi.org/10.1101 /2020 Our test results revealed that there are five potential breaks in both intercept and trend, 21 March, 27 March, 2 April, 12 April, and 20 April. The first break on 21 March reflects the initial impact of the measures, the rest of the breaks seem to reflect the growing impact of the different policy measures that are beginning to kick. We specified alternative break dates both in the intercept and trend, all other possible intercept and trend breakpoints perform substantially worse in terms of getting white noise residuals and fit. More simply, we estimated model (1') to determine whether measures are taken by Turkish authorities The estimation results with intercept and trend break with 2 and 5 are given in Table 1 . was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which this version posted May 4, 2020. Table 1 , Column 2 presents the result with two breakpoints and column 3 with five breakpoints. Information criteria indicate that the model with 5 breaks is better than the model with 2 breaks. Furthermore, the model with 2 breaks has the serial correlation in the residuals. Furthermore, Figures 3 and 4 show the log of confirmed cases and the fitted values from the model with 2 breaks and 5 breaks, respectively. It is obvious from these figures that the model with 5 breaks fits better than the model with 2 breaks. All rights reserved. No reuse allowed without permission. was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Our analysis confirms a pronounced slowdown in the growth of confirmed Covid-19 infections in Turkey starting from 21 March 2020. While the growth rate has slowed down significantly, the All rights reserved. No reuse allowed without permission. was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint (which this version posted May 4, 2020. . https://doi.org/10. 1101 /2020 number of cases is still doubling approximately every 11 days during the last week of April. Due to substantial delays between new infections and their measurement in statistics, we could see further effects from the government measures in the near future. Corona measures need to be taken carefully and timely. During the implementation of these measures, it is necessary to avoid practices that will cause the general public to panic. If this cannot be done, social distance measures will be neglected and measures to prevent the spread of the disease will be harmed. The spread of Covid-19 is central to the public health and the economy, therefore, we will follow the development of the confirmed growth rates closely. Estimation of a change point in multiple regression models Estimating and testing linear models with multiple structural changes Computation and Analysis of Multiple Structural Change Models Critical Values for Multiple Structural Change Tests Public Health Measures to Slow Community Spread of Coronavirus Disease Measuring the impact of the German public shutdown on the spread of Covid-19 Coronavirus COVID-19 global cases by the centre for systems science and engineering The Incubation Period of Coronavirus Disease 2019 (COVID-19) From Publicly Reported Confirmed Cases: Estimation and Application Incubation period and other epidemiological characteristics of 2019 novel Coronavirus infections with right truncation: A statistical analysis of publicly available case data Daily situation report on the SARS-CoV-2 in Germany Prediction on Covid-19 epidemic for different countries: Focusing on South Asia under various precautionary measures