Credit risk in start ups: a literary approach

What does Honoré de Balzac and credit risk have in common? Well, maybe nothing. But wait…reading his bio one finds that Balzac’s inexperience and lack of capital caused his ruin in many trades. He borrowed money from family and friends to build a printing business only to give it to a friend (who finally made them successful). He carried the debt for many years.

In the novel “The rise and fall of Cesar Birotteau”, Balzac describes the life of a Parisian perfumer who achieves success in cosmetics business, but becomes bankrupt due to property speculation.

In one of the chapters Balzac (himself worked in a law office) explains the bankruptcy laws as they existed in his time and how they were abused by dishonest businessmen who wanted to escape their debts.

“The rise and fall of Cesar Birotteau” is one of the novels of the “Scènes de la vie parisienne” in the series “La Comédie Humaine”. This book is, in fact, a collection of interlinked 91 novels, stories and essays depicting French society between 1815 and 1848. Balzac explained his aim in compiling all the works was to understand “social species” in the way biologists would analyse “zoological species”.

Ok- let’s move now to credit risk. Credit risk is originated when typically a bank gives money to a company, based on some analysis the bank performs using more or less data and more or less a mathematical framework, to come up with the probability of the company defaults on its debt. The less data the company could offer (for example an early stage tech start up), the less accurate the probability of default assessment will be. That’s why traditional banks do not want to lend to startups at early stages.

But, at the end of the day, credit risk is the output of Balzac’s social species behaviours in conducting business: how they make decisions, how they design and manage incentives, how they react taking into account the reaction of others…the difference between startup A success and startup B failure might be startup B founders team was not able to work with one another (too much ego?). If company B owed € 5 M to the bank, then the bank will have to write off company’s B debt regardless all the funny-and-complex-machine learning-credit risk models used to assess company’s B creditworthiness during the loan origination process.

In my view, reading Balzac and many other writers is the best way to be inspired and to learn how to cope with many of the current challenges in startup financing- mainly when data is scarce and information asymmetry reigns.

Let me go now through two credit risk related topics where literature may give a bit of light, or at least food for thought. These are topics that every time and again disturb my risk analyst mind (I have many more but they will hopefully be topics for next posts).

ACCEPT OR DENY A LOAN TO AN EARLY STAGE START UP

Analysing a loan for a startup is one of the most complex things one comes up in banking industry. First there is the fact that by nature, an early stage startup typically lacks enough cash flow to even pay the interest of a loan. In this phase instead, venture capital enters injecting equity, which is a more stable funding source. Some traditional banks devise innovative loan add-on features that make them more interesting from a risk management point of view. But this is a topic for another post.

Some weeks ago, I had the opportunity to hear a Silicon Valley venture capital founder in a dinner speech. She was asked how she identified good from bad projects.

Cicero, the Roman philosopher, in his book “De divinatione” (On divination) offers two types of methods to see into the future. One is based in skills and interpretation (oracles, astrology and augury). The other is based in inspirations (e.g. dreams).

After more than 2000 years of Cicero and dozens and dozens of books in credit risk, prediction science and finance, I think Cicero’s way to predict is still the essence of all the methodologies. Today, roughly speaking, by Cicero’s “skills and interpretation” we would mean quantitative models and by “inspiration” we would mean expert judgment.

In deciding whether we lend to a startup, and in general to thin file applicants, data is incomplete, scarce and unstructured. The analysis then shifts to expert judgment, rather than quantitative models. Henry Marsh, a well known neurosurgeon and best seller author, famously said he needed 3 months to learn to operate, but 30 years to learn when not to operate. Just replace “operate” by “granting a loan” and you will have the risk management dilemma.

By the way, the Silicon Valley founder (more than 25 years of expertise) answered that when analyzing whether to invest in a company, she went through the business plan, the size of the market, did the number crunching, she interviewed with the founder, etc but many times, in her most successful companies, the decision to invest came up without her performing a rational assessment or applying a mathematical formulae. She said the founder seemed a sober person, a visionary, the team was aligned, the mission was clear or there was a product fit for the market, but to me, what she wanted to say is that she simply invested by Cicero-like inspiration/expert judgment. And this judgment is very difficult to teach, unlike the, say, CAPM or DCF models.

PROJECTS COMING FROM INCUBATORS/ACCELERATION PROGRAMMES

Startups incubators are like schools for startups. Cohorts of entrepreneurs work in an office space together and go through a series of tutorials and mentorship programmes over a 3 month period to see what comes out.

The considered number 1 is Y Combinator, having facilitated the success of Dropbox, Airbnb, Reddit and many more. Founded in 2005, it has invested in more than 1500 startups worth a total of $ 80 bn. Every year receives 13000 applications, out of which picks out between 200 and 240. An injection of capital follows.

Where the startup was incubated is an important factor when deciding to grant a loan. A startup coming from an incubator with an excellent performance provides with more comfort (i.e. lower risk of failure) when it comes the loan application process. Incubators (and VCs) most valuable capital is the reputation for backing successful startup companies. So the positive brand signalling matters, as Scott Kupor says in his book Secrets of Sandhill Road.

My problem when analysing incubators’ performance is (and this is also applicable to venture capital funds): which is the most successful one? The majority of incubators show the ratio of successful projects (whatever you call successful) over the total projects as a key performance indicator… So what follows is: if a startup accelerated by one of the top 10 successful incubators asks for a loan, then the likelihood of this startup to becoming a kind of new Google is very high. But wait…..what about the failed projects also accelerated by those very incubators?

In the Black Swab, Nassim Taleb uses the term “silent evidence” to describe this phenomenon. He introduces it with a story from Cicero:

“Diagoras, a nonbeliever in the gods, was shown painted tablets bearing the portraits of some worshippers who prayed, then survived a subsequent shipwreck. The implication was that praying protects you from drowning.

Diagoras asked, “Where are the pictures of those who prayed, then drowned?”

Those “drowned believers” are silent evidence, i.e failed startups also incubated by Top 10 incubators that don’t show up in key performance metrics. We do have to take them into account when assessing the incubator/venture capital performance.

As many readers of this article sure know, venture capital returns does not follow a normal-gaussian distribution. That’s why the average return is not a robust metric for performance. Returns in venture follows a power law distribution: there are 10–20% projects where you will win 100 or 1000 times the original investment, 50–60% where you will lose nearly all the money and 20–30% where you probably will win 1–2 times the original money invested.

From a risk management point of view, then it would be better to forecast a venture loan/fund performance based not in expected losses (based on the mean of the distribution) but in unexpected losses (based on the volatility of the distribution).

That said, lending activity for startups will always be a risky one, full of errors due to incomplete information, market volatility and so forth.

May the banks find comfort in David Hume, who once said: “the most important task is to recognize the absolute impossibility of any certainty”. Two centuries later, Jorge Luis Borges added: “to accept errors is not to contradict chance; it is to corroborate it”.

Note: Please if you had a business challenge with your start up, something disturbing your mind during night dreams and were helped by reading a book , or a quote from an author, please shared it to me. I’m collecting all them and will then post all your useful insights. We could all benefit from your readings!!! email: roman.weissmann@gmail.com