The data and the damage done

Let’s start with a story.

Jane left college at 18 — her parents couldn’t afford to send her to University and she didn’t want to saddle herself with debt that it would take years to pay off. She works in her local supermarket. Some weeks are good and she gets loads of shifts. Other weeks are bad and she doesn’t get as many as she needs. Due to the unpredictability of her main income she tends to keep a couple of part-time jobs too and currently does some shifts at a petrol station and a bar in town. Rent and other monthly living expenses need to be paid and sometimes the unpredictability of her earnings puts her in a temporarily tough financial position.

She’s very money conscious, checks her bank account daily and is very wary about debt. She avoids credit and only wants to take the minimum she needs when essential. Unforeseen short-term expenses put her significantly behind on her budget. She goes to her bank — the bank she’s used her entire life — for an overdraft and is refused because she doesn’t have a regular monthly paycheck. Although not keen on credit cards as she fears slipping into the never-never, she applies for one anyway and is denied for the same reason. She tries the bank for a loan of £250, but they won’t lend such a small amount. Exasperated, she turns to a payday lender. It’s going to cost her £310 to pay it back in one month but she doesn’t see any alternative.

She’d asked for extra shifts at the bar knowing she needed the money. Unfortunately, both weekend shifts she got were quite quiet and as the part-timer she got sent home early. That puts her £50 further behind in her budget. When the payment to the payday lender falls due she can’t make it and is immediately slapped with a £15 charge. Luckily she gets some extra shifts at the supermarket and she can pay both the loan and charge off.

John, who was at school with Jane, decided that although his parents couldn’t afford to send him to University student loan debt isn’t ‘real’ and once he’d graduated he’d snag a well paying job and pay it all off. He’s now graduated and moved to London. As a graduate he’s got a £1,000 overdraft and is living paycheck by paycheck. It’s a decent paycheck but essential outgoings are high and then there are the social pleasures of graduate life in London to indulge. He rarely looks at his bank balance — too dull and it’ll work out alright (particularly if he gets that pay rise next year). He wants to spruce up his flat and is surprised when his debit card doesn’t work on the Ikea website. Without thinking he clicks on the VISA card offer from his bank, signs up online and is immediately approved. He buys his furniture and then when the month end comes around he’s not got enough in his account to pay off the debt and his rent so his direct debit fails. He gets a nasty email from the bank a few days later, but with a message in it that if he switches his VISA card to a monthly minimum fee option he’ll be fine. He logs on, one click and everything is sorted.

Now, looking at these two friends it’s quite obvious which one is the best with money. In fact, many studies have shown that contrary to received wisdom the poorer someone is the better they are at managing their finances.

But if we looked at these two individuals through the prism of traditional credit data we’d see the following:

Now, our point here is not the arbitrary value of Jane’s and John’s score but that they have nothing to do with whether an individual is honest, trustworthy or ‘good with money’. Their implied wealth largely dictates their initial score which in turn largely dictates the products available to them. These products have an overwhelming effect on the future path of their credit scores.

Obviously wealth is a key characteristic to measure when you’re lending large amounts over a long period, but for short-term credit we believe its main purpose is to justify exorbitant fees. With the right short-term credit product there is no reason why Jane should have an implied default profile that’s any worse than John’s. In fact, we believe with the right product it should be much better.

Think about it this way: John undeniably has the ‘safety valve’ of higher income and better job prospects but risk is a complex thing and in many ways he has bigger financial pitfalls he can fall into, partly as a result of the amount of credit and ease by which he can access it and the size of his overheads. A lender to him is exposed to considerably more downside than with Jane if something goes wrong in John’s economic life. Jane on the other hand has much more modest credit needs, lower overheads and, over the medium term, can adapt quicker than John to find cash to pay off outstanding debts. There’s a deep and perverse tautology at play here around wealth and already having credit. Step outside the tautological framework of wealth and existing credit indicating future creditworthiness and it is obvious that data delivers us a very malformed credit environment.

Back to our story, and in the current credit-data reality Jane would now find herself firmly in what we term the ‘debt trap’. This is a pernicious cycle whereby an individual faces an ever constricting window of access to fairly priced mainstream financial services. The newly acquired focus on affordability and vulnerability from regulators and the concern sector works in tandem with the data to set the trap. Increasingly the consumer find themselves at the mercy of parasitic businesses and products that are structurally designed to further disenfranchise them. In Jane’s case, one necessary and cautious dalliance with high cost short-term credit and the damage is done…

In modern society many people on the lower end of the income scale have an increasingly volatile net income flow. If these same people lose access to our shared humongous balance sheet and the flexibility it provides then, no matter how good with money they are, they are going to have to utilise products that us middle classers turn our noses up at. It’s simply a matter of surviving to the next paycheck. It doesn’t mean they don’t understand money and need financial literacy assistance [1].

How did we get here?

Forty years ago credit wasn’t brokered using data. It used to be that if you wanted a loan you made an appointment with your local bank manager. He’d come around to your house for a cup of tea, take a look at your furniture, the state of your shoes etc. and knowing your status in the local community he’d make a decision. Credit was and always had been intermediated via social interaction — the relationship, and all the potentially awkward things that could entail, was the information fed into the underwriting process.

However, big finance didn’t like this. It didn’t scale. And when global financialisation exploded on to the scene we needed a new way to appraise consumer credit. Credit scores were born and our deeply infatuated and unquestioning love affair with data began.

Now, this ‘datafication’ has been an incredible boon for the global middle classes. It’s enabled 110% mortgages, explosion in the availability of credit and a collapse in its cost. However, for the lower middle class and working class it’s often been a disaster. When you’re an asset light or low income individual then, from a financial transaction perspective, your character is your most important asset. In the past you could pledge this asset in person by dealing directly with the individual who would make the credit decision. They could take the measure of you.

With algorithms and numbers taking the place of more instinctive but complex human decisions the lowest earners in society became increasingly disenfranchised. We have no rosetinted spectacles on the historic social aspect of credit as it could be intrusive and even grating or embarrassing for borrowers. However, it was in that potential social awkwardness that the user was able to form and transact a valuable social collateral that modern data has no way of replicating[2]. Modern credit data only sees a blend of how wealthy the user is and how successful they have been in recent years at keeping up to date with credit repayments. Even if you are 100% upstanding you are going to miss some payments if your income is low and/or variable…and that means you get double-whammied on lack of wealth and missed payments in the data-credit world.

Now, if you talk to any credit reference bureau they will tell you that their scores translate directly to an individual’s character and this would seem logical, right? I mean, if a person has missed a payment to a payday lender with all their scary fees and interest payments they’ve got to be a pretty untrustworthy person, right? Well, re-read the story at the top of this post….

You see, the problem is that credit is highly reflexive.

Credit is reflexive

George Soros, the billionaire hedge fund manager and philanthropist, has often articulated a conceptual framework called reflexivity that he’s deployed to devastating effect during his career. It enabled him to anticipate the 2008 financial crisis and predict and react to events better than almost anyone else.

Although complex its core concept can be explained in very simple terms, in his own words:

I can state the core idea in two relatively simple propositions. One is that in situations that have thinking participants, the participants’ view of the world is always partial and distorted. That is the principle of fallibility. The other is that these distorted views can influence the situation to which they relate because false views lead to inappropriate actions. That is the principle of reflexivity. For instance, treating drug addicts as criminals creates criminal behavior. It misconstrues the problem and interferes with the proper treatment of addicts. As another example, declaring that government is bad tends to make for bad government.

His philosophical insight is as true in consumer credit as it is in drug treatment and politics. The situation surrounding a credit transaction is as important as the outcome of that credit transaction. Treat a consumer as a credit pariah or cretin and that is what you will create.

If you’re a consumer with a bank overdraft and a wide plethora of credit products at your fingertips you’re not more trustworthy than an individual whose only access to credit is through a payday lender. But traditional credit data would tell you that you are.

Having spent a good part of the last two years analysing the various data points traditional credit data scoring provides, it seems to us that this data and the way it is used tends to primarily tell you how wealthy someone is. Which if they’re coming to you for an emergency cash advance you already know the answer to[3].

The solution — more data??

Many have witnessed the shortcomings that traditional credit data has when used to extend small ticket credit to people in the lower income levels of our society. Concerned onlookers, lenders and financial regulators all see the issues.

But the problem is that in almost every case their answer seems to be more data. “If we can just get that data set that tells us the transaction history of their bank account we’ll know whether they can afford the loan”, “if we can just get that data set that tells us precisely where they work and how much they earn we’ll be able to better judge their credit worthiness” etc. This is just mega data dumbness. We have a terribly tendency to believe that numbers exactly match or predict reality and we are suckers for people and firms that sell this idea to us. It is why we managed to make a terrible global financial hole back in 2008 as we believed the data around complex credit instruments was accurate and real. Of course, events in the global financial crisis illustrated just how detached reality and the data can be.

In 2008 we discovered that data zealot traders in banks had been given so much freedom to create data-driven alternate realities that they nearly sunk the world. Now we’re discovering that fintechers — the supposed disruptors of these broken banks — are just as algo-addled and divorced from reality. Unfortunately, most onlookers, including legislators and the regulator, share this data dumbness and have an unhealthy compulsion to believe that more data and poorly conceived underwriting and algorithmic processes are the answer. Wonky data, starting from a place that is wrong and not aligned to the reality it seeks to model is not the foundation to build upon, particularly with more data[4]. Yet again regulators and legislators have been suckered into believing this data fantasy of the world. Wake-up! You are a party to the disenfranchisement and harm caused to an increasing percentage of your population.

Having access to more data points doesn’t address the need to go with the grain of the reflexive nature of credit. People are not a point on a risk spectrum, they are a distribution of potential risks. Risk outcome for any one transaction is a complex function of the context within which the transaction took place and the continuing relationship between borrower and lender thereafter.

Data providers of course charge for access to their data. So more data pushes up the cost of credit for a consumer who can least afford the increase in cost. If data is used ‘responsibly’ in the current system then the minority of people who are simultaneously relatively wealthy enough but in need of a payday loan effectively pay for the cost of all the people that fail the data check in the price they pay for their credit. The industry has responded to a combination of price caps and higher costs with the ‘innovation’ of high cost instalment loans (see ‘Paydaynomics’). This is a mind blowingly counterproductive situation but it is what Stella Creasey and various other concerned individuals and groups have ended up celebrating as a success and the FCA seems to implicitly support as a solution.

The solution — bigger data??

And then we have our fintech brethren. You’ve got to love these guys. All sweet faced and bright eyed twenty somethings with VC money to burn to relearn the law of diminishing returns.

That’s harsh. On the face of it their solution would seem to make sense. Focus on analysing the character of an individual by pulling together disparate non-financial datasets from social media etc.

The problem is that getting access to these datasets is getting harder and harder as individuals, wising up to identity theft issues, are becoming increasingly suspicious of providing personal information to third party service providers. This leaves them trying to scrape sites for this information and then trying to match up applicants with data points. With a huge number of locations now that consumers leave their social data, gaining and then maintaining access to these locations is not cheap. Anyone with experience of a capital intensive, low margin business knows how this story ends.

We think the final death knell in the ‘big data’ solution is the fact that people have already been doing it. Hamburg based Kreditech have been lending for 3 years now using their proprietary big data solution. If this big data was the solution to the problems highlighted above this would give them an immediate cost advantage so you’d expect them to be considerably cheaper than a lender using traditional credit scoring, right? Well…

So what is the solution?

We’ll give you a hint. It isn’t data.

We’re not going to lay out our answer here as we’re a commercial venture with intellectual property to protect. However, we hope that this post and its predecessors can help you see a little through our eyes and encourage you to question and, hopefully, doubt the consumer credit system. We think it is broken. It is most obvious for lower income consumers but it is very complacent to believe this is just a bottom 20% problem. We think it’s growing beyond that now and is hunting down lower middle class consumers, keen to use data to show they’re more risky than they are and should be excluded from the mainstream credit balance sheet.

Credit has and always will be an innately social transaction. Based on trust and social bonds; a credit relationship has to be formed before the transaction takes place and maintained after it completes. When seen this way it’s bonkers to think it can be reduced down to 0s and 1s.

The challenge is doing this in a scalable way…

If you want to know more, want to tell us we’re wrong or want to get involved then drop us a comment below or email me on tb@leanonafriend.com.

Tom

Footnotes

[1] Remember that next time you see a finance business touting its ability to help teach poor people how to better manage their money.

[2] Legislation and regulatory guidance focuses quite heavily on not embarrassing a customer around credit issues. We agree with the spirit of these rules and avoiding unnecessary embarrassment but we do question whether there is an excessive and counterproductive de-stigmatisation of credit and that this is most harmful to lower income segments of society.

[3] This appears to us to be a sort of perverse systemic form of reflexivity. And, except for the profits of the subset of companies that benefit from the credit disenfranchisement of consumers, this seems to be a wholly undesirable societal outcome. Finance is always political and if this corporate violence was being perpetrated on “decent home-owning hard-working aspirational people” there would be a revolution.

[4] See our first post ‘ A disrupters view of UK payday’ for further discussion on how the data and reality are mismatched.