Avoiding Lockdowns and Covid-19 Deaths without Waiting for Vaccines

How to end the epidemic faster, by introducing mandatory use of privacy protecting contact tracing apps (e.g. CoronaWarnApp) and widespread free testing. #NoCovid4dot0

Various restrictions have been imposed on personal freedoms in order to curb the spread of Covid-19 until the widespread vaccination of the population can be achieved. The cost of these restrictions is enormous, affecting not only the economy, but also personal well-being, individual freedoms and the education of a generation. In this piece I argue that large scale automated contact tracing can reduce the need to use lockdowns to a minimum, while respecting the privacy of citizens.

We will go over the epidemiological, economic and political advantages of widespread automated contact tracing compared to other strategies and give concrete policy proposals for implementing it.

If you are fearful of or opposed to mandatory contact tracing app usage, be brave, read on and confront yourself with another viewpoint. If you aren’t, get explanations and arguments in support of it so you can share these ideas and just maybe we can end the crisis of our times a few months earlier.

Epidemiological models show automated contact tracing is one of the most effective strategies to reduce the spread of Covid-19

There are to my knowledge 5 ways to reduce the spread of Covid-19:

  • Reducing Contacts
  • Improving Hygiene and Protection
  • Vaccination
  • Testing and Quarantining
  • (Automated) Contact Tracing

In this section, we analyze their individual and combined theoretical efficacy based on epidemiological models. Automatic contact tracing will be shown to be among the most effective strategies in isolation and when combined with other strategies. In case you already know how effective contact tracing is, already know the basics of epidemiology or get easily tired by looking at plots and simulation results, feel free to skip down to the economic and political arguments below.

How the modelling works

In this section we will show plots that simulate epidemics using models that split the population into groups or states (susceptible, exposed, infected, recovered, vaccinated, dead). We will use two modeling approaches:

  • A simple one that looks at averages and transition rates between groups. (Differential equations)
  • And an agent based approach in which each person is represented by an agent which undergoes state transitions with a certain probability, that allows more realistic modelling.

These are two variants of the standard compartmentalized models in epidemiology also known as SIR-models. For each day they compute how an idealized epidemic would proceed given probabilities that certain things (contacts, infections, death, etc.) happen. Parameter values were used that are in the range of possible values as indicated by credible sources for Covid-19. Nevertheless, these simulations are not meant to be predictive, but rather to show the relative effect of various strategies to combat an epidemic. If you are especially interested in the simulation details or want to know references or exact parameters, feel free to read, leave comments and play with the code.

Baseline Simulation

In the following you will see many plots such as this where the percentage of people in each state is shown over time. We show both the result of the agent based model (solid lines), the variance between simulations (shaded area) and where applicable the result of the equivalent differential model (dashed line). The models are run with ten to hundreds of thousands of agents due to computational constraints, but the general dynamics hold for larger populations.

Here the same data is plotted with a logarithmic y-axis so you can better see the deaths rising together with the infections.

These plots show what happens if no action is taken, and assumes recovery means immunity. We see that after 100 days, which is relatively short, the epidemic is over. Since almost the entire population is infected, the mortality rate (IFR 1.0%) applies to the whole population. In a country like Germany, that would correspond to 800 000 deaths, as well as many more serious cases that may cause long term symptoms. In addition, the mortality would increase significantly if the hospital system were overwhelmed, but this is not modeled here or throughout.

Reducing Contacts and Improving Hygiene

By reducing how many contacts people have and improving hygiene, the number of people a sick person will infect on average decreases. The number that models this is the basic reproduction number, R0. In these plots we reduce R0 by reducing the number of daily contacts for each simulation. Since this dynamic depends on R0, we would get similar results if we instead reduced the infection risk during each contact.

We see that reducing contacts makes the epidemic proceed more slowly. However, it does not alter the number of dead or infected by orders of magnitude, as long as R0 is significantly above 1. Reducing the number of contacts to one is basically equivalent to living in a permanent lockdown and is thus unrealistic.


Here we vary the daily vaccination rates, measured by the amount of time until the whole population is vaccinated. Since many cases are undiagnosed and governments seem to plan to vaccinate even those who have recovered, we model that vaccine doses spent on those who have recovered are wasted. In other terms, there is no transition from Recovered to Vaccinated in the model.

Vaccinations and the resulting immunity obviously stop the spread of a disease, if a high enough vaccination and recovery rate is achieved. However, as expected our model shows that even with high vaccination rates, many people die if it is not accompanied by other measures, since an unimpeded exponential spread of the disease is much faster than the linear growth of the vaccinated population.


In this simulation a lockdown strategy is implemented where the number of contacts varies depending on the diagnosed infection rate in the population. Values for the incidence rates for entering and exiting lockdowns are chosen similar to those that have been set in Germany. Here hard lockdowns are conceptually stay-at-home orders while medium lockdowns are a combination of hygiene and physical distancing. Lockdowns are modeled by reducing the number of daily contacts for the entire population. Note that the x axis is much larger for this simulation and spans 10 years, since the dynamics have slowed.

Since lockdowns are very effective at keeping peak infection rates low, we additionally plot the same data with a logarithmic y axis so we can see if the algorithmic lockdown strategy works as intended.

We see that the strategy indeed enters a lockdown every time the incidence goes above the predefined levels, much more quickly and with much more resolve than any government would.

Lockdowns in and of themselves just delay an epidemic. This protects hospitals from reaching overcapacity, but in the absence of vaccines or improved treatment, a similar number of people die as in the baseline model (where overcapacity is not modeled) although with much higher economic cost.

Testing and Quarantining

In the following simulations we test a certain percentage of the infected agents every day and if they test positive put them in quarantine where they have much fewer contacts. Here we vary both the ratio of undiagnosed cases to diagnosed cases as well as the testing speed. Testing speed is modeled by both the mean time after which an infected agent gets tested and the time it takes for them to receive their test results.

We see that as long as testing is slow, even if almost everyone is tested, which is unrealistic, the epidemic is not stopped. And we again have mortality similar to the baseline. Only if testing is fast and widely applied does it have a significant impact in slowing down transmission.

Thus the focus should be on widespread and easy availability of testing and on making sure test results are quickly received. However testing alone only stops the epidemic if everyone is quickly tested positive which is likely not feasible.

Contact Tracing

In these simulations we vary the use and speed of contact tracing. In each of these simulations, the undiagnosed ratio is 2.5 (40% diagnosed). In addition, for the final simulation we use recursive contact tracing, where not only contacts of those that have been tested positive are put in quarantine but also all contacts of those contacts, in order to get ahead of the spread.

We see that both testing only and slow (manual) contact tracing merely slows down the spread somewhat versus the baseline model but the epidemic only ends once most of the population has been infected. Both automated fast contact tracing variants show a wide range of outcomes with some of them completely eliminating the disease and a median mortality much lower than the baseline. Recursive contact tracing performs the best, but does put a significant percentage (median 5%, max 20%) of the population in quarantine (Testing out of quarantine early is only partially modeled). In this model no social distancing has been used and only 40% of infections were diagnosed, yet the epidemic is almost stopped in its tracks.

When modelling the other strategies, we had to select unrealistically beneficial parameters for them to stop the epidemic, but here we were able to choose parameters which may be achievable. For further reading on contact tracing and its effectiveness read this great article from April 2020 by Tomas Pueyo. Recent scientific Nature papers also show the effectiveness of the recursive contact tracing, also known as bidirectional and backwards contact tracing.

Summarizing the Individual Measures:

Physical distancing and lockdowns work to reduce the spread of the disease but need to last indefinitely. Vaccinations take too long to produce and distribute to stop a novel epidemic. Testing and quarantining is only effective if one can be tested quickly, very high diagnosis rates are achieved and result notification is fast, i.e. automated. Contact tracing can stop an epidemic on its own as long as it is fast, automated and uses recursive tracing.

Current Lockdown and Vaccination Strategy

As far as I can tell the basic government strategy in Germany and many other countries is to rely on lockdowns and some testing to delay the spread of the disease until enough people have been vaccinated. Here we model such a combined lockdown, vaccination and testing strategy. In this model vaccinations only start after one year once the vaccine is available. We model both a pessimistic scenario with slow testing and vaccination rates and an optimistic scenario, where both rates are high. We average the lockdown state across all simulation runs, which is why they can also have values other than 50% or 100% although in each individual simulation these were the only options.

We see that, although the strategy works and saves many lives, it takes one and a half to two years until the epidemic has ended and severe lockdowns were imposed for a large portion of that time.

Proposed Mandatory Automated Contact Tracing, Lockdown and Vaccination Strategy

Here we model a combined strategy with the same parameters as above, with the addition of automated recursive contact tracing.

We see from the simulations that both scenarios spend less time in strict lockdown and more lives are saved than in the plots without contact tracing above.

The gains are less obvious in the left plot, when fewer cases are diagnosed, since then of course fewer cases are traced. Still, this difference translates into 40,000 (0.05%) lives saved. In addition, one third less time (140 vs 210 days) is spent in hard lockdown, which assuming a daily lockdown cost of a few billion euros, saves a couple hundred billion euros or thousands of euros per person in Germany.

In the right plot of the optimistic scenario, in which 40% of cases are diagnosed, mortality and infection rates are very low and no time is spent in a hard lockdown. Here around 60 000 lives are saved and again thousands of euros per person are not spent, compared to the simulation without automated contact tracing.

Thus we have shown that, according to our model, automated contact tracing can allow us to begin opening months earlier and avoid the most severe restrictions entirely while saving more lives compared to the current strategy.

Why automated contact tracing is more effective than manual contact tracing

Most health authorities understand that only contact tracing really stops a pandemic in its tracks without vaccination, and that is why the early focus was on it. However, it seems that after investing large sums and man-hours to manual contact tracing, many governments have had to give up on it as caseloads rose anyway. This is understandable, since contact tracing by human contact tracers is tedious work and is sometimes simply not feasible due to the number and type of contacts.

In contrast, privacy protecting automated contact tracing has not received as many resources, even though it is much more effective, since it:

  • enables tracing contacts whose identities are unknown to the infected person.
  • is much faster, since the manual effort of finding out people’s phone numbers or addresses is not needed and there is no lag between the time the test is confirmed positive and when the health authority can begin its work.
  • does not break down when a certain incidence level is reached. Manual contact tracing fails when there is simply no longer enough staff to do the contact tracing.
  • preserves more privacy, since one does not literally tell the government the names of people one have been in contact with.

Now that we have described why automated contact tracing is one of the most promising and underutilized strategies to combat the Covid-19 pandemic, I will try to explore and counter some of the reasons why democratic governments have not attempted to enforce it and why there may be widespread opposition among the population against it.

The exposure notification framework is privacy preserving, encrypted, anonymous, decentralized and does not use your position.

Many people say they do not want to use contact tracing apps due to privacy concerns or fears of centralized tracking of people's movements. These fears are however unfounded and stem from a general fear of technology and tracking rather than an understanding of the privacy preserving exposure notification framework (ENF).

The ENF works by transmitting and receiving random numbers to and from nearby bluetooth enabled smartphones and storing them locally for later comparisons with the random numbers associated with infected people. Here I would like to describe why the ENF is privacy protecting and is not a tool with which anyone can be surveilled.

The exposure notification framework is:

  • Privacy preserving — Contacts are only recorded as random ids (numbers) which change frequently. A user has no way of knowing whom a random id belongs to, only the day the contact happened and its approximate duration. If you have very few contacts this may be enough to guess who the contact was but the key point is you are not learning anything you did not already know. You already knew you had contact with this person and hopefully they would have told you they have tested positive anyway. It would just have taken longer and not been anonymous.
  • Position-less — Your position is never recorded or used in the ENF. (Older Android versions have an unfortunate OS permissions system which gives the impression that position data is perhaps being used, it is not. It is possible to turn this permission on for the OS so the ENF works but revoke it for every other app. This issue has likely cost many lives but cannot be changed without updating all OSs for devices that are no longer supported by their manufacturers, which the OEMs and Google are unwilling or unable to do.)
  • Anonymous — You are never identified by name, only as a sequence of random numbers that changes every 10–20 minutes.
  • Decentralized — Your random ids are never uploaded unless you were diagnosed with Covid-19. Only the changing random numbers associated to the time windows in which infected people were infectious are uploaded to a central server. All crosschecking of which infected random ids a user had contact with happens on the user’s phone. This means the central server does not even know what contacts between random ids happened.
  • Encrypted — The few bits of nonrandom information, such as the approximate distance to the contact are encrypted.
  • Temporary — The history of anonymous random locally stored ids is deleted after two weeks.

Many data protection advocates support the ENF because a lot of thorough work has been done to make it protect your privacy. It protects your privacy so well that actually enforcing a usage mandate would be more difficult than enforcing a mask mandate and would likely involve having you show your phone to the authorities.

Yes, the technology behind the ENF is difficult to understand, but that alone is not a reason to eschew it. Sometimes we should just trust that there are problems technology can help solve.

But hasn’t automated contact tracing using the exposure notification framework failed?

No, it hasn’t even really been tried since usage is so low. Automated contact tracing is a quadratic problem since both users have to use the app for the contact to be detected. This means at 30% usage (optimistic estimate for Germany) only 9% of contacts are traced and it has not had an obvious impact.

However after the mandate is introduced, usage may initially be around 70%, with the result that 50% of contacts of diagnosed cases would be traced. In younger age groups that have the most contacts, usage and contact tracing rates could approach 100%. Once recursive contact tracing is enabled, a similarly significant portion of previously undiagnosed cases will be traced and diagnosed.

In countries that have used mandatory automated contact tracing systems (Singapore, China, Korea) they have been effective at keeping Covid-19 at bay, although their contact tracing systems are very intrusive and Orwellian in nature. They can thus be compared to this proposal in epidemiological effectiveness but not with regards to their protection of privacy.

Economics and Cost

Here we would like to go through the economics of each of the proposed strategies. We will only be speaking in general terms since I was unable to find good sources for the exact costs and economic modelling is even more difficult than epidemiological modelling. The relative costs can however be approximately judged using the known cost of existing measures.

Improving hygiene with face masks, disinfectants and other manufactured goods is relatively cheap. Although the cost of these basic goods is low, over time the recurring cost may add up. In addition, as we saw in the beginning of the pandemic, the delay in the ramp up in production can lead to shortages.

Vaccinations are incredibly cheap, costing only a few tens of euros per person, once every few years and allows a path back to normal once population immunity has been reached. Because of this, no cost should be spared in accelerating vaccine development and production. The problem is that they still involve the physical world, leading to production difficulties, susceptibility to mutation as well as unfair distribution of the limited supply.

Testing a significant portion of the population on a weekly or monthly basis could incur a cost of up to a few hundred euros per person per year, depending on the cost of an individual test. As the epidemic continues, this cost will likely decrease as tests get cheaper. However, testing is critical to guiding epidemic policy no matter what strategy policymakers choose.

Physical distancing is almost free to implement, simply involving the adherence to rules. However there are many high indirect costs, such as the overall reduced economic activity due to capacity limits and inefficiencies caused by remote work.

As we have all experienced, lockdowns with business and school closures are incredibly, inconceivably costly, up to the tens of thousands of euros per person year. All of the treasure and man hours for which debt is being collected, could have been used for addressing other issues important to each of us such as: combating climate change, fighting poverty or improving education worldwide, or just for living comfortable lives. In addition, the educational deficits caused by school closures incur severe lifetime economic costs and, more importantly, decrease the general well being and health of those affected.

Digital, automated contact tracing using cell phones most people (Germany: 80%) already own is however almost free. Although the development effort may be significant, it is unlikely to cost more than a few euros per person. Since it only involves downloading and installing an app, the cost of “production” is almost nonexistent and its “production” velocity infinite. Partially because of its low cost policymakers may not believe how incredibly valuable it is in combating epidemics, compared to alternative strategies. As we have shown here, it should be one of the most important tools in the policy chest, given its low cost per life saved.

Thus, if we were to order the mitigation measures in terms of cost, we would likely get an order similar to:

Automated Contact Tracing < Vaccines < Hygiene < Testing < Physical Distancing < Lockdowns

Where we have on the left measures that only have small one-time cost, followed by measures that have small but repeated cost and then finally measures that have severe continuous costs. This should thus likely also be the order in which measures are applied.

Political Considerations

I know the proposal for required installation of a contact tracing application is controversial, to put it mildly.

In addition, politicians have promised that contact tracing frameworks would not become mandatory. This was a mistake, but we should allow them and ourselves to revisit past decisions and not allow past promises to hamper overcoming one of the greatest challenges of our time.

Always, but especially in times of crisis, governments must weigh certain basic rights against each other. In this case, the right to a healthy life and freedom of movement and enterprise. When it comes to privacy preserving contact tracing apps however, the only right being infringed upon is the right to only have apps of your choice on your phone. Does such a right exist? (Why can Youtube or iMessage not be uninstalled if it does?)

Maybe you are fearful of a requirement to install an app that acts as an anonymous beacon and want to believe your gut feeling that this means we will start living in a dystopian society where your every movement is tracked. I invite you to compare such a requirement to other restrictions imposed by freely elected governments as well as the tracking many of us freely submit to and rethink whether it really moves us in that direction.

Our democratically elected governments have decided it is acceptable to:

  • Force you to stay in your home and restrict your freedom of movement.
  • Require you to tell health authorities the names, times and places the contacts occurred.
  • Limit the number of people you have contact with.
  • Require your business to close and thus restrict your freedom of enterprise.
  • Save your communications metadata for a certain period of time, to fight crime. Data retention policies that allow governments to access the identity, time, duration, recipient, device and sometimes location associated with communications are in effect in most european democracies and the US.
  • Let businesses track your every click and touch of a mouse in every application or website you use. This is the industry standard: if you are using a computer or smartphone, you are being tracked and you accepted being tracked, when you accepted the terms of service. (Thankfully in the EU because of GDPR they do have to ask you every time for non anonymous data, but not for the anonymous data.)

All of this is ok and acceptable, yet it is not supposed to be ok for the government to require you to install and use a privacy preserving contact tracing app to save lives and the economy?

It is already relatively common for governments to require the use of technical means to protect people from potential danger. It has long been a requirement to use seat belts or lights when using cars and bikes. Today mask wearing is required in many places. So I genuinely ask:

Which right does being required to use a privacy preserving automatic contact tracing system in an epidemic infringe upon, in unacceptable ways?

Policy Proposals to End the Epidemic

As shown, mandatory automated contact tracing is the most effective and cheapest way to slow an epidemic, according to our models and economic reasoning. Some changes to current testing policy and contact tracing app usage and implementation are however required to achieve these theoretical results in reality.

1st day proposals:

  1. Legally require contact tracing app usage in public transport, stores and workplaces, similar to mask requirements. Users that carry a signed waiver that they do not have a capable smartphone are exempt.
  2. Implement sporadic checks that users are using the app and not breaking quarantine.
  3. Make sure all types of testing (PCR and rapid) are free and available to all in test centers. Only impose limits if test capacity is being reached.
  4. Start considering warnings given by the contact tracing app as formal quarantine notices.
  5. Highlight the privacy preserving nature of the contact tracing app.

1st month proposals:

  1. Enable recursive keys contact tracing from people who have been in contact with positive cases. Force Google and Apple to implement this in the Exposure Notification Framework. This is what actually stops the epidemic since it gets ahead of the spread.
  2. Make sure all testing is required to be integrated and automated within the app. Testing without automated contact tracing is only of limited epidemiological value, given large rates of undiagnosed infections. Worrying about false positives should play only a limited role in testing and contact tracing while the epidemic is still raging.
  3. Order cheap capable phones or other bluetooth enabled devices for the 20 percent of people who do not have capable smartphones yet.
  4. Spare no cost to improve automation, app development and test integration. Use the entire power of the government to accelerate this development and remove any impediments to this automation.
  5. It may be necessary to create an app for use by authorities to enable them to check that nonusers have signed the no smartphone ownership waiver and verify the authenticity of contact tracing app installations, since fake contact tracing apps may emerge to help people circumvent the install requirement. This does not reduce privacy, compared to the authorities checking that mask mandates are being followed and taking names when they are not.

Of course, these proposals are just meant as a starting point for actual implemented policy and many variations and improvements upon this plan are conceivable.


As we have seen automated contact tracing is one of the most effective ways to stop an epidemic, is one of the cheapest strategies to implement and infringes the least upon your rights. In addition, concrete measures were proposed to implement mandatory usage of a privacy protecting contact tracing app.

Especially for countries where vaccinations are not proceeding quickly, I hope the policies proposed here can lead to reduction of loss of life, improve the economy, maintain freedom of movement and promote well-being. Even if it proves to be too late to affect policy changes in this pandemic, I hope this piece may be used to reduce the impact of future pandemics and inform future policymakers.

Please share if you think the ideas here are worth spreading and comment if you have more to add. I would like to hear what anybody who disagrees with the arguments and conclusions made here has to say and am open to constructive discussions.

We can still outsmart this thing!