Pandemic: why use adjusted data?

Chart of the week

Figures from the French national public health agency, Santé publique France, show that the nation’s Covid-19 situation has significantly improved in the last few days: the daily number of new cases has slowed significantly, the number of new hospitalisations is falling and the positivity rate is decreasing.

The improvement is visible in most regions and particularly noteworthy in Ile-de-France, which includes Paris and its neighbouring departments. Ile-de-France is France’s richest and most populous region, accounting for approximately one third of GDP and 20% of the population. Here, we take a closer look at whether we can deduce that the improvement in the Ile-de-France health situation is a direct result of lockdown.

Our analysis

Due to a strong seasonality effect, this question is difficult to answer based purely on the raw data provided by Santé publique France. Slower data transmission at weekends results in systematically lower numbers for Saturdays and Sundays followed by higher numbers on Mondays due to the catch-up effect.

To overcome this problem and identify a trend, the data are smoothed weekly. The problem is that the resulting figures reflect not only the current pandemic situation, but also the past. To adjust the data series shown in the above chart, we tried stripping out the recurring weekly gaps. This approach, which is similar to the technique used to seasonally adjust economic data, put the focus back on the current situation and enabled us to better identify when the pandemic peak was reached.

Strikingly, the situation in Ile-de-France appears to have been improving before the second lockdown was imposed on 30 October. According to our calculations, a peak in new Covid-19 cases in Ile-de-France was reached on 24 October, six days before the new lockdown started and well before its effects could materialise. Assuming a 3–5-day virus incubation period, immediate testing and very fast results, it takes at least a week for new cases to show up in the figures. Meanwhile, the smoothed data show a peak on 28 October, which is the day the new lockdown arrangements were announced and four days later than the adjusted data.

We can conclude that the pattern of lockdown followed by trend reversal in Ile-de-France masks a more complex reality. Perhaps the pre-lockdown restrictions would have been enough to reverse the trend. The arrival of rapid antigen tests, which could mechanically reduce the number of cases by substituting PCR tests, cannot explain the decline in new cases after 24 October because they were only made available in pharmacies on 31 October and in modest quantities. What is for sure is that analysing seasonally adjusted data series can help better interpret changes and detect peaks earlier. As such, it a valuable decision-making tool.


See also :


The opinion expressed above is dated 19th November 2020 and is liable to change.


This document is not pre-contractual or contractual in nature. It is provided for information purposes. The analyses and descriptions contained in this document shall not be interpreted as being advice or recommendations on the part of Lazard Frères Gestion SAS. This document does not constitute an offer or invitation to purchase or sell, nor an encouragement to invest. This document is the intellectual property of Lazard Frères Gestion SAS. LAZARD FRERES GESTION – a simplified joint stock company with share capital of €14,487,500 – Paris Trade and Companies Registry No. 352 213 599. 25, RUE DE COURCELLES – 75008 PARIS, FRANCE