Something for Nothing – Higher Levels of Service with Marginal Increases in Cost

A Free Lunch - Let the Government Do All the Work



In our first blog post we explained the challenges associated with implementing peer to peer insurance on Ethereum, proposed a set of criteria an insurance DAO (Decentralized Autonomous Organization) would need to meet and then made the statement that supplemental unemployment insurance provided a unique opportunity to fulfill all of our requirements. Dynamis will be one of the first practical attempts to issue insurance on Ethereum by offering a supplemental UI policy. It will be using LinkedIn as both a reputation system to establish an online identity and as an oracle to establish the validity of claims.

The goal of this blog post is to try to give people a framework of understanding the difference between traditional insurance administered by traditional corporations and insurance administered by a DAO. To understand how different an approach this is as a point of reference we need to consider how a traditional supplemental unemployment insurance company would perform the tasks of claims verification and validation.

The US government provides UI benefits and to do so it takes the following actions:

  1. Establishes the start of a claim – processes the claimant’s application and verifies dates of previous employment.
  2. Evaluates a claim’s legitimacy, awards or denies a claim.
  3. Monitors the job search of the claimant and validates their continuing eligibility for receiving claim payments.
  4. Establishes the end of a claim – W4 forms received from next employer.
  5. Penalizes claimants for fraud resulting in overpayment of a claim.

What private insurers do:

  1. Request proof that a claimant has been granted UI payments by the government.
  2. Continue to pay supplemental UI payments so long as the government is paying on a claim.

Offering supplemental UI is a brilliant system because verifying a claimant’s status, awarding claims, and combating fraud for supplemental UI benefits are externalized costs to private insurance providers. When realizing this I almost felt stupid that I was trying create a parallel solution because traditional insurance providers are already getting a free lunch so any improvements would have to be pretty radical to trump the existing system. I said ‘almost’ because I suspected that a DAO could solve the problem of claims verification and fraud mitigation multiple different ways and all of them should cost the same. This is because the operation of a DAO does not require any fulltime staff. How variable could costs be to run different claims verification and fraud mitigation methods on the blockchain? Would different methods result in more expensive off blockchain infrastructure? Intuitively it seems that costs would be roughly equivalent but I won’t know until I have more data. Perhaps one solution is more labor intensive to code but beyond that initial cost both solutions should cost the same for the DAO to execute. Traditional providers may be able to save money by relying on the government but so long as a DAO can solve the problem of claims verification and fraud mitigation the solution should not consider the cost to operate as a factor.

Granted if one solution had a great deal more fraud than the other then one solution would cost more. But if you had several different methods to mitigate fraud and they all did so with various degrees of success they probably all cost the DAO about the same to execute regardless of a methods effectiveness. The revelation that costs of operation are no longer directly tied to the quality of an outcome forces the reevaluation of traditional methodologies.

In the end relying on the government does have drawbacks and there are potential benefits to trying a new way of awarding, validating and paying claims. The more you try to form a comparison between an insurance DAO and a traditional insurance institution the more you realize how different both approaches to providing insurance are. Although you could try to create a DAO that underwrites policies, collects premiums, evaluates and pays claims similarly to traditional insurance institutions it wouldn’t work well because the trust model is very different. Centralized systems use centralized trust models. In the traditional supplemental UI model the government is the ultimate source of factual authority. This saves traditional institutions money and effort.

Here are some benefits the traditional system has going for it:

  1. Did I mention it is nearly free for insurance companies?
  2. Employers and claimants are required by law to cooperate with the government’s verification process.
  3. Fraud deterrence is also free because the government puts measures in place to penalize people who commit soft fraud and can easily enforce penalties thus reducing fraudulent overpayments.
  4. Having claimants submit proof of government awards and providing training to existing policy holders who can validate the claimant is receiving government payments is trivial.

Some downsides to relying on the government:

  1. You are limiting yourself to national governments, every national government has different policies and processes for awarding and validating claims.
  2. Issues with overpayments:
    • In the US the government pays out two weeks at a time and they do not prosecute for fraud when overpayment is one week or less.
    • The percentage of claims which end on the first and third week of the month are overpaid by at least one week (non-fraud overpayment). If claims are on average 20 weeks or less this could result in an overpayment of 5% or more.
    • The U.S. Labor Department estimates that 11% of all payments made on a claim were overpayments.
  3. The honor system that the US government uses requires claimants to self-report on the progress of their job search. There is no easy way to validate if claimants are doing any work at all to find their next job and inevitably some people may have given up looking while still collecting.
  4. Many in the cryptocurrency community (potential initial policy holders) fundamentally disagree with centralized trust models and would choose not to participate if we used claim award and validation methods employed by traditional UI companies.

The government is not that motivated to reduce costs and overpayments as no one is going to put it out of business if it remains inefficient. By privatizing the award and validation process could traditional insurers gain additional cost savings? As inefficient at paying out claims the government may be a free lunch for insurance companies is still preferable. Claims investigators are expensive overhead. An investigator’s access to information is incomplete relative to the government because no one is mandated by law to cooperate with an insurance company’s verification of claims. As they research a claim it’s conceivable that an investigator’s incomplete information as to the claimant’s status might result in more fraud and overpayments and not less.

Relying on government validation of unemployment is more cost efficient than private validation?

Inconceivable!

Inconceivable!

Well not exactly. The safer option is the free one at least you know how much it costs. Traditional insurers would have to be crazy to come up with a “better” system. I can just imagine sitting down with an insurance executive and explaining how Dynamis works. “The government overpays claims? So what. I can break out a calculator and tell you what a supplemental UI premium payment needs to be to make a profit; why bother finding a ‘better’ system.”

Decentralized Trust Models Being Fundamentally Different


Dynamis is different because it does not use a centralized trust model not because it cannot but because it chooses not to. It might be significantly less effort to try and copy how traditional supplemental UI providers evaluate claims. It would also be a great deal more boring. My initial motivation for starting this project was to do something different, not to replicate things as they currently work. I knew nothing about traditional insurance let alone anything related to supplemental UI. After listening to Vitalik give a talk on Schelling points I was inspired and I just wanted to create something that used Schelling points and I didn’t care what it was. It just so happened that it turned out to be insurance. I didn’t attempt to find a better way to underwrite policies or evaluate claims. No. I just wanted to make a Schelling point smart contract do something cool. And I also loved LinkedIn and I wanted to use it somehow. Since LinkedIn is a social media site which ties peoples online identities to their employment history I figured I would just try supplemental UI to see if it might work.

Other than the fact that Dynamis is heavily reliant on LinkedIn, a centralized database which serves as a reputation system and as an oracle, the model that it uses to underwrite policies and award claims is different than a centralized trust model in the following ways:

  1. Underwriting policies and awarding claims is not managed by any central authority but decided collectively by a subset of policy holders (peers). They work in coordination with the DAO executing the insurance organizations corporate policy. Hopefully this results in less bias against individual claimants.
  2. The entire underwriting and evaluation process is transparent and auditable.
  3. The method by which a claim is validated and paid is not primarily determined by any central authority but through the exercise of a claimant’s ‘social capital’ which is really the positive support the claimant receives from their own social network.

Dynamis uses a decentralized trust model because this better relates to a DAOs unique set of capabilities. DAOs can publish transparently with no additional overhead but they also cannot survive without the blockchain, a transparent publishing database. For DAOs to use HITs (Human Intelligence Tasks) to automate the process of underwriting policies and evaluating claims seems good because it might remove human biases but DAOs also cannot survive without leveraging human intelligence to solve certain tasks. Even if relying on the government to perform all claim verification and fraud deterrence was more cost efficient it would be irrelevant.

Centralized trust models clash with the unique strengths a decentralized platform such as Ethereum offers to developers and end-users. If you want to create a platform for insurance with a centralized trust model you don’t need Ethereum but that doesn’t mean you can’t use Ethereum. You need to know your target audience and appeal to them by leveraging the unique strengths and assets a DAO is capable of. Centralized trust models do not appeal to the Ethereum community just like soy sauce flavored Kit-kat (gross Japan) is never going to be popular in the US. As it turns out reliance on the government probably isn’t more cost efficient because computationally speaking the cost difference between running different variants of code are probably negligible.

A Free Lunch Another Way – Humans as Intelligence Resources


When you award and validate claims the lazy way through the government you end up having no incentive to actually help people find a job. Dynamis provides mutual peer to peer intermediary free insurance and as an added bonus the claims validation process facilitates people’s job search. The goal of traditional UI companies is to find the easiest and least expensive way of sustaining the payment of claims by charging the right price for premiums which includes a profit margin for the corporation. This is how insurance companies stay alive. Traditional companies could spend additional money to offer job search assistance but that would mean higher premiums and less competitive advantage on the basis of price. If the policy holder is paying more and the assistance is a value added benefit this may give a traditional intuition a competitive advantage on the basis of service. Traditional companies using traditional methods have always been forced to make these types of tradeoffs; DAOs however may be able to offer increasingly higher levels of service with marginal increases in cost. In essence they can get something for very little (perhaps nothing) as they break new services down into fragmentary tasks which can be automated in coordination with humans (cost efficient intelligence resources) who perform HITs to augment the logic of the DAO (cost efficient administrative engines). But this theory has yet to be tested.

Hopefully data will become available in the next year which will shed some light on these theories and perhaps DAOs which perform similar tasks but with different code and trust models will be created which will allow us to analyze the relationship between a DAO’s operating costs and the level of service it provides users. This will certainly give us some real insights into what we can hope to expect in the future from smart contract resources. At the expense of sounding repetitive: the fundamental difference between Dynamis and traditional insurance institutions is that the insurance DAO uses a decentralized trust model and this model produces different results which were previously unexpected. The way it awards and validates claims is not easier or less expensive in terms of operating overhead than using the government as a single trusted oracle, it is simply different. It does require more effort on behalf of the claimant because Dynamis will not be using a claimant’s employers to determine the start or end of a claim. You no longer have a single centralized authority as an oracle which means you will need for the claimant to exercise their social capital to validate claims and release claim payments. Put another way the claim validation oracle is the collective reporting done by members within a claimant’s social network. The endorsement of a claimant’s employment status from a claimant’s social network has the ability to release claim payments (monetary capital). Using social networks in this way allows DAOs to integrate social media transactions into the mechanism of financial transactions which authorize insurance claim payments. The unexpected byproduct of claim validation being performed within the context of a social network is that insurance DAOs help people find jobs.

Now I hope you can begin to see how radically different the insurance DAO is from traditional insurance companies. How can you really compare them? In five years if they both offer the same exact services who do you think will win? If traditional insurers wish to survive they will need to do more than focus on competitive pricing of premiums that provide corporations a small profit margin on top of paying out claims and administrative costs. In our example above even when claim verification and fraud mitigation were free for insurance companies they were also free for the DAO using a completely different system. A DAOs administrative overhead is always going to be lower than traditional insurers which means that so long as they wish to compete on price alone they are always going to lose. Now insurers are forced to answer difficult questions. Helping people find a job costs money. Do you want to pass that cost onto policy holders? Value added services are nice but they price you out of traditional markets. The insurance DAO however will soon take away the ability for traditional institutions to compete merely on the basis of the pricing of premiums.

This gives you a brief glimpse into the coming upheaval that smart contracts will bring to nearly every market in the coming years. The possibilities are endless? Well so long as the Luddites don’t try to bring down the network I suppose. The power of smart financial systems using new distributed publishing technologies does have a limit but trying to imagine what it might be is both exciting and unnerving.

In the coming blog posts we will explore how human participants interact with the DAO and some of the technical details surrounding how the processes of underwriting policies, awarding claims and validating a claimant’s status are automated.

Your fellow cost efficient intelligence resource – J.D.